To improve vision-based methods for understanding how people interact with their physical environment, we introduce a multi-view video and body-contact sensing dataset designed to capture dynamic human activities that involve interactions with the physical environment. The dataset includes activities such as parkour, physical training, and gym exercises, characterized by frequent body-environment contact. The proposed dataset includes 780 K images across 120 K pose sequences from 7 subjects. Each subject is captured by 6 synchronized third-person cameras, a single egocentric camera, and multiple contact sensors worn on the body. Using our proposed dataset, we benchmark state-of-the-art vision-based body contact models and show significant limitations in existing methods. Furthermore, we benchmark existing human pose estimation methods on our dataset and show that they fail under significant occlusion caused by close interactions with the environment, which indicates that our dataset can also be used to further develop pose estimation models to be more robust during interaction with the environment. To facilitate better human pose estimation from video, we introduce and evaluate a video-based human contact detection model that outperforms existing imagebased methods, underscoring the potential improvements from integrating contact information into pose estimation models. See the project page at: https://yohanshin.github.io/bodycontact4d.github.io/

Accurate three-dimensional (3D) human motion tracking is essential for immersive augmented reality$(A R)$and virtual reality (VR) applications, allowing users to engage with virtual environments through realistic full-body avatars. Achieving this level of detail, however, is challenging when the driving signals are sparse, typically coming only from upper-body sensors, such as head-mounted devices and hand controllers. To address this challenge, we propose EgoMDM (Egocentric Motion Diffusion Model), an end-to-end diffusion-based framework designed to reconstruct full-body motion from sparse tracking signals. EgoMDM models human motion in a conditional autoregressive manner using a unidirectional recurrent neural network, making it well-suited for real-time applications. By embedding local-to-global translation, forward and inverse kinematics, and foot-contact detection within the diffusion framework, EgoMDM achieves seamless, end-to-end motion synthesis, effectively reducing artifacts like foot sliding and ground penetration. Additionally, EgoMDM is conditioned on the user's body scale, allowing it to generalize across a diverse population and produce consistent avatar shapes over time. In our extensive experiments on the AMASS motion capture dataset, EgoMDM achieves state-of-the-art performance in both motion tracking accuracy and synthesis quality, demonstrating its robustness and adaptability across various human motion scenarios. Furthermore, EgoMDM significantly outperforms the existing models when tested on real signal inputs, highlighting its robustness and applicability to the real-world data. See the project page at: https://yohanshin.github.io/egomdm.github.io/

Understanding how humans interact with objects is key to building robust human-centric artificial intelligence. However, this area remains relatively unexplored due to the lack of large-scale datasets. Recent datasets focusing on this issue mainly consist of activities captured entirely in controlled lab environments, and contact annotations are mostly estimated using threshold clips. We introduce Contact4D, a multi-view video dataset for humanobject interaction that provides detailed body poses and accurate contact annotations. We use a flexible multiview capture system to record individuals performing furniture assembly tasks and provide annotations for human detection, tracking, 2D/3D pose estimation, and groundtruth contact. Additionally, we propose a novel processing pipeline to extract accurate hand poses even when they are severely occluded. Contact4D consists of$2 M$images captured from 19 synchronized cameras across 350 video sequences, spanning diverse environments, varioius furniture types, and unique subjects. We evaluate existing methods for human pose estimation and human-centric contact estimation, demonstrating their inability to generalize to our dataset. Lastly, we fine-tune a pretrained MultiHMR model on Contact4D and observe an improved performance of 56.6% body MPJPE and 26.4% hand MPJPE in scenarios under severe self-occlusion and object occlusion. Code and data are available at https://jyuntins.github.io/Contact4D.

Feedback is essential for learning a new skill or improving one's current skill-level. However, current methods for skill-assessment from video only provide scores or compare demonstrations, leaving the burden of knowing what to do differently on the user. We introduce a novel method to generate actionable feedback (AF) from video of a person doing a physical activity, such as basketball or soccer. Our method takes a video demonstration and its accompanying 3D body pose and generates (1) free-form expert commentary describing what the person is doing well and what they could improve, and (2) a visual expert demonstration that incorporates the required corrections. We show how to leverage Ego-Exo4D's [29] videos of skilled activity and expert commentary together with a strong language model to create a weakly-supervised training dataset for this task, and we devise a multimodal video-language model to infer coaching feedback. Our method is able to reason across multi-modal input combinations to output full-spectrum, actionable coaching—expert commentary, expert video retrieval, and expert pose generation—outperforming strong vision-language models on both established metrics and human preference studies.

We introduce HOIGPT, a token-based generative method that unifies 3D hand-object interactions (HOI) perception and generation, offering the first comprehensive solution for captioning and generating high-quality 3D HOI sequences from a diverse range of conditional signals (e.g. text, objects, partial sequences). At its core, HOIGPT utilizes a large language model to predict the bidrectional transformation between HOI sequences and natural language descriptions. Given text inputs, HOIGPT generates a sequence of hand and object meshes; given (partial) HOI sequences, HOIGPT generates text descriptions and completes the sequences. To facilitate HOI understanding with a large language model, this paper introduces two key innovations: (1) a novel physically grounded HOI tokenizer, the hand-object decomposed VQ-VAE, for discretizing HOI sequences, and (2) a motion-aware language model trained to process and generate both text and HOI tokens. Extensive experiments demonstrate that HOIGPT sets new state-of-the-art performance on both text generation (+2.01% R Precision) and HOI generation (-2.56 FID) across multiple tasks and benchmarks.

Robotic grasping is a cornerstone capability of embodied systems. Many methods directly output grasps from partial information without modeling the geometry of the scene, leading to suboptimal motion and even collisions. To address these issues, we introduce ZeroGrasp, a novel framework that simultaneously performs 3D reconstruction and grasp pose prediction in near real-time. A key insight of our method is that occlusion reasoning and modeling the spatial relationships between objects is beneficial for both accurate reconstruction and grasping. We couple our method with a novel large-scale synthetic dataset, which comprises 1M photo-realistic images, high-resolution 3D reconstructions and 11.3B physically-valid grasp pose annotations for 12K objects from the Objaverse-LVIS dataset. We evaluate Zero-Grasp on the GraspNet-1B benchmark as well as through real-world robot experiments. ZeroGrasp achieves state-of-the-art performance and generalizes to novel real-world objects by leveraging synthetic data. https://sh8.io/#/zerograsp

While recent advancements in camera-based 3D object detection demonstrate remarkable performance, they require thousands or even millions of human-annotated frames. This requirement significantly inhibits their deployment in various locations and sensor configurations. To address this gap, we propose a performant semi-supervised framework that leverages unlabeled RGB-only driving sequences - data easily collected with cost-effective RGB cameras - to significantly improve temporal, camera-only 3D detectors. We observe that the standard semi-supervised pseudo-labeling paradigm underperforms in this temporal, camera-only setting due to poor 3D localization of pseudo-labels. To address this, we train a single 3D detector to handle RGB sequences both forward and backward in time, then ensemble both its forwards and backwards pseudo-labels for semi-supervised learning. We further improve the pseudo-label quality by leveraging 3D object tracking to infill missing detections and by eschewing simple confidence thresholding in favor of using the auxiliary 2D detection head to filter 3D predictions. Finally, to enable the backbone to learn directly from the unlabeled data itself, we introduce an object-query conditioned masked reconstruction objective. Our framework demonstrates remarkable performance improvement on large-scale autonomous driving datasets nuScenes and nuPlan.

Parametric body models offer expressive 3D representation of humans across a wide range of poses, shapes, and facial expressions, typically derived by learning a basis over registered 3D meshes. However, existing human mesh modeling approaches struggle to capture detailed variations across diverse body poses and shapes, largely due to limited training data diversity and restrictive modeling assumptions. Moreover, the common paradigm first optimizes the external body surface using a linear basis, then regresses internal skeletal joints from surface vertices. This approach introduces problematic dependencies between internal skeleton and outer soft tissue, limiting direct control over body height and bone lengths. To address these issues, we present ATLAS, a high-fidelity body model learned from 600k high-resolution scans captured using 240 synchronized cameras. Unlike previous methods, we explicitly decouple the shape and skeleton bases by grounding our mesh representation in the human skeleton. This decoupling enables enhanced shape expressivity, fine-grained customization of body attributes, and keypoint fitting independent of external soft-tissue characteristics. ATLAS outperforms existing methods by fitting unseen subjects in diverse poses more accurately, and quantitative evaluations show that our non-linear pose correctives more effectively capture complex poses compared to linear models.

We present Ego-Exo4D, a diverse, large-scale multimodal multiview video dataset and benchmark challenge. Ego-Exo4D centers around simultaneously-captured egocentric and exocentric video of skilled human activities (e.g., sports, music, dance, bike repair). 740 participants from 13 cities worldwide performed these activities in 123 different natural scene contexts, yielding long-form captures from 1 to 42 minutes each and 1,286 hours of video combined. The multimodal nature of the dataset is unprecedented: the video is accompanied by multichannel audio, eye gaze, 3D point clouds, camera poses, IMU, and multiple paired language descriptions—including a novel “expert commentary” done by coaches and teachers and tailored to the skilled-activity domain. To push the frontier of first-person video understanding of skilled human activity, we also present a suite of benchmark tasks and their annotations, including fine-grained activity understanding, proficiency estimation, cross-view translation, and 3D hand/body pose. All resources are open sourced to fuel new research in the community. https://ego-exo4d-data.org/

We introduce Ego4D, a massive-scale egocentric video dataset and benchmark suite. It offers 3,670 hours of daily-life activity video spanning hundreds of scenarios (household, outdoor, workplace, leisure, etc.) captured by 931 unique camera wearers from 74 worldwide locations and 9 different countries. The approach to collection is designed to uphold rigorous privacy and ethics standards, with consenting participants and robust de-identification procedures where relevant. Ego4D dramatically expands the volume of diverse egocentric video footage publicly available to the research community. Portions of the video are accompanied by audio, 3D meshes of the environment, eye gaze, stereo, and/or synchronized videos from multiple egocentric cameras at the same event. Furthermore, we present a host of new benchmark challenges centered around understanding the first-person visual experience in the past (querying an episodic memory), present (analyzing hand-object manipulation, audio-visual conversation, and social interactions), and future (forecasting activities). By publicly sharing this massive annotated dataset and benchmark suite, we aim to push the frontier of first-person perception.

We present SimXR, a methodfor controlling a simulated avatar from information (headset pose and cameras) ob-tained from AR / VR headsets. Due to the challenging view-point of head-mounted cameras, the human body is often clipped out of view, making traditional image-based ego-centric pose estimation challenging. On the other hand, headset poses provide valuable information about overall body motion, but lack fine-grained details about the hands and feet. To synergize headset poses with cameras, we control a humanoid to track headset movement while analyzing input images to decide body movement. When body parts are seen, the movements of hands and feet will be guided by the images; when unseen, the laws of physics guide the controller to generate plausible motion. We design an end-to-end method that does not rely on any intermediate representations and learns to directly map from images and headset poses to humanoid control signals. To train our method, we also propose a large-scale synthetic dataset created using camera configurations compatible with a commercially available VR headset (Quest 2) and show promising results on real-world captures. To demonstrate the applicability of our framework, we also test it on an AR headset with a forward-facing camera.

We present Ego-Exo4D, a diverse, large-scale multi-modal multiview video dataset and benchmark challenge. Ego-Exo4D centers around simultaneously-captured ego-centric and exocentric video of skilled human activities (e.g., sports, music, dance, bike repair). 740 participants from 13 cities worldwide performed these activities in 123 different natural scene contexts, yielding long-form captures from 1 to 42 minutes each and 1,286 hours of video combined. The multimodal nature of the dataset is un-precedented: the video is accompanied by multichannel audio, eye gaze, 3D point clouds, camera poses, IMU, and multiple paired language descriptions-including a novel “expert commentary” done by coaches and teachers and tailored to the skilled-activity domain. To push the frontier of first-person video understanding of skilled human activity, we also present a suite of benchmark tasks and their annotations, including fine-grained activity understanding, proficiency estimation, cross-view translation, and 3D hand/body pose. All resources are open sourced to fuel new research in the community.

While recent depth completion methods have achieved remarkable results filling in relatively dense depth maps (e.g., projected 64-line LiDAR on KITTI or 500 sampled points on NYUv2) with RGB guidance, their performance on very sparse input (e.g., 4-line LiDAR or 32 depth point measurements) is unverified. These sparser regimes present new challenges, as a 4-line LiDAR increases the distance between pixels without depth and their nearest depth point sixfold from 5 pixels to 30 pixels compared to 64 lines. Ob-serving that existing methods struggle with sparse and variable distribution depth maps, we propose an Affinity-Based Shift Correction (ASC) module that iteratively aligns depth predictions to input depth based on predicted affinities between image pixels and depth points. Our framework enables each depth point to adaptively influence and improve predictions across the image, leading to largely improved results for fewer-line, fewer-point, and variable sparsity settings. Further, we show improved performance in domain transfer from KITTI to nuScenes andfrom random sampling to irregular point distributions. Our correction module can easily be added to any depth completion or RGB-only depth estimation model, notably allowing the latter to perform both completion and estimation with a single model.

We propose G-HOP, a denoising diffusion based generative prior for hand-object interactions that allows modeling both the 3D object and a human hand, conditioned on the object category. To learn a 3D spatial diffusion model that can capture this joint distribution, we represent the human hand via a skeletal distance field to obtain a representation aligned with the (latent) signed distance field for the object. We show that this hand-object prior can then serve as generic guidance to facilitate other tasks like reconstruction from interaction clip and human grasp synthesis. We believe that our model, trained by aggregating seven diverse real-world interaction datasets spanning across 155 cate-gories, represents a first approach that allows jointly generating both hand and object. Our empirical evaluations demonstrate the benefit of this joint prior in video-based reconstruction and human grasp synthesis, outperforming current task-specific baselines.

We present a universal motion representation that encompasses a comprehensive range of motor skills for physics-based humanoid control. Due to the high dimensionality of humanoids and the inherent difficulties in reinforcement learning, prior methods have focused on learning skill embeddings for a narrow range of movement styles (e.g. locomotion, game characters) from specialized motion datasets. This limited scope hampers their applicability in complex tasks. We close this gap by significantly increasing the coverage of our motion representation space. To achieve this, we first learn a motion imitator that can imitate all of human motion from a large, unstructured motion dataset. We then create our motion representation by distilling skills directly from the imitator. This is achieved by using an encoder-decoder structure with a variational information bottleneck. Additionally, we jointly learn a prior conditioned on proprioception (humanoid's own pose and velocities) to improve model expressiveness and sampling efficiency for downstream tasks. By sampling from the prior, we can generate long, stable, and diverse human motions. Using this latent space for hierarchical RL, we show that our policies solve tasks using human-like behavior. We demonstrate the effectiveness of our motion representation by solving generative tasks (e.g. strike, terrain traversal) and motion tracking using VR controllers.

We propose a new visual hierarchical representation paradigm for multi-object tracking. It is more effective to discriminate between objects by attending to objects’ compositional visual regions and contrasting with the background contextual information instead of sticking to only the semantic visual cue such as bounding boxes. This compositional-semantic-contextual hierarchy is flexible to be integrated in different appearance-based multi-object tracking methods. We also propose an attention-based visual feature module to fuse the hierarchical visual representations. The proposed method achieves state-of-the-art accuracy and time efficiency among query-based methods on multiple multi-object tracking benchmarks.

Developing autonomous vehicles that can safely interact with pedestrians requires large amounts of pedestrian and vehicle data in order to learn accurate pedestrian-vehicle interaction models. However, gathering data that include crucial but rare scenarios - such as pedestrians jaywalking into heavy traffic - can be costly and unsafe to collect. We propose a virtual reality human-in-the-loop simulator, JaywalkerVR, to obtain vehicle-pedestrian interaction data to address these challenges. Our system enables efficient, affordable, and safe collection of long-tail pedestrian-vehicle interaction data. Using our proposed simulator, we create a high-quality dataset with vehicle-pedestrian interaction data from safety critical scenarios called CARLA-VR. The CARLA-VR dataset addresses the lack of long-tail data samples in commonly used real world autonomous driving datasets. We demonstrate that models trained with CARLA-VR improve displacement error and collision rate by 10.7% and 4.9%, respectively, and are more robust in rare vehicle-pedestrian scenarios.

We present Human to Humanoid (H2O), a reinforcement learning (RL) based framework that enables real-time whole-body teleoperation of a full-sized humanoid robot with only an RGB camera. To create a large-scale retargeted motion dataset of human movements for humanoid robots, we propose a scalable "sim-to-data" process to filter and pick feasible motions using a privileged motion imitator. Afterwards, we train a robust real-time humanoid motion imitator in simulation using these refined motions and transfer it to the real humanoid robot in a zero-shot manner. We successfully achieve teleoperation of dynamic whole-body motions in real-world scenarios, including walking, back jumping, kicking, turning, waving, pushing, boxing, etc. To the best of our knowledge, this is the first demonstration to achieve learning-based real-time whole-body humanoid teleoperation.

We present a method for controlling a simulated humanoid to grasp an object and move it to follow an object's trajectory. Due to the challenges in controlling a humanoid with dexterous hands, prior methods often use a disembodied hand and only consider vertical lifts or short trajectories. This limited scope hampers their applicability for object manipulation required for animation and simulation. To close this gap, we learn a controller that can pick up a large number (>1200) of objects and carry them to follow randomly generated trajectories. Our key insight is to leverage a humanoid motion representation that provides human-like motor skills and significantly speeds up training. Using only simplistic reward, state, and object representations, our method shows favorable scalability on diverse objects and trajectories. For training, we do not need a dataset of paired full-body motion and object trajectories. At test time, we only require the object mesh and desired trajectories for grasping and transporting. To demonstrate the capabilities of our method, we show state-of-the-art success rates in following object trajectories and generalizing to unseen objects. Code and models will be released.

To predict future trajectories, the normalizing flow with a standard Gaussian prior suffers from weak diversity. The ineffectiveness comes from the conflict between the fact of asymmetric and multi-modal distribution of likely outcomes and symmetric and single-modal original distribution and supervision losses. Instead, we propose constructing a mixed Gaussian prior for a normalizing flow model for trajectory prediction. The prior is constructed by analyzing the trajectory patterns in the training samples without requiring extra annotations while showing better expressiveness and being multi-modal and asymmetric. Besides diversity, it also provides better controllability for probabilistic trajectory generation. We name our method Mixed Gaussian Flow (MGF). It achieves state-of-the-art performance in the evaluation of both trajectory alignment and diversity on the popular UCY/ETH and SDD datasets. Code is available at https://github.com/mulplue/MGF.

Video transformers are slow to train due to extremely large numbers of input tokens, even though many video tokens are repeated over time. Existing methods to remove uninformative tokens either have significant overhead, negating any speedup, or require tuning for different datasets and examples. We present Run-Length Tokenization (RLT), a simple approach to speed up video transformers inspired by run-length encoding for data compression. RLT efficiently finds and removes `runs' of patches that are repeated over time before model inference, then replaces them with a single patch and a positional encoding to represent the resulting token's new length. Our method is content-aware, requiring no tuning for different datasets, and fast, incurring negligible overhead. RLT yields a large speedup in training, reducing the wall-clock time to fine-tune a video transformer by 30% while matching baseline model performance. RLT also works without training, increasing model throughput by 35% with only 0.1% drop in accuracy. RLT speeds up training at 30 FPS by more than 100%, and on longer video datasets, can reduce the token count by up to 80\%. Our project page is at rccchoudhury.github.io/projects/rlt.

Understanding how humans interact with each other is key to building realistic multi-human virtual reality systems. This area remains relatively unexplored due to the lack of large-scale datasets. Recent datasets focusing on this issue mainly consist of activities captured entirely in controlled indoor environments with choreographed actions, significantly affecting their diversity. To address this, we introduce Harmony4D, a multi-view video dataset for human-human interaction featuring in-the-wild activities such as wrestling, dancing, MMA,and more. We use a flexible multi-view capture system to record these dynamic activities and provide annotations for human detection, tracking, 2D/3D pose estimation, and mesh recovery for closely interacting subjects. We propose a novel markerless algorithm to track 3D human poses in severe occlusion and close interaction to obtain our annotations with minimal manual intervention. Harmony4D consists of 1.66 million images and 3.32 million human instances from more than 20 synchronized cameras with 208 video sequences spanning diverse environments and 24 unique subjects. We rigorously evaluate existing state-of-the-art methods for mesh recovery and highlight their significant limitations in modeling close interaction scenarios. Additionally, we fine-tune a pre-trained HMR2.0 model on Harmony4D and demonstrate an improved performance of 54.8% PVE in scenes with severe occlusion and contact. “Harmony—a cohesive alignment of human behaviors." Code and data are available at https://jyuntins.github.io/harmony4d/.

We propose SolePoser, a real-time 3D pose estimation system that leverages only a single pair of insole sensors. Unlike conventional methods relying on fixed cameras or bulky wearable sensors, our approach offers minimal and natural setup requirements. The proposed system utilizes pressure and IMU sensors embedded in insoles to capture the body weight’s pressure distribution at the feet and its 6 DoF acceleration. This information is used to estimate the 3D full-body joint position by a two-stream transformer network. A novel double-cycle consistency loss and a cross-attention module are further introduced to learn the relationship between 3D foot positions and their pressure distributions. We also introduced two different datasets of sports and daily exercises, offering 908k frames across eight different activities. Our experiments show that our method’s performance is on par with top-performing approaches, which utilize more IMUs and even outperform third-person-view camera-based methods in certain scenarios.

Kalman filter (KF) based methods for multi-object tracking (MOT) make an assumption that objects move linearly. While this assumption is acceptable for very short periods of occlusion, linear estimates of motion for prolonged time can be highly inaccurate. Moreover, when there is no measurement available to update Kalman filter parameters, the standard convention is to trust the priori state estimations for posteriori update. This leads to the accumulation of errors during a period of occlusion. The error causes significant motion direction variance in practice. In this work, we show that a basic Kalman filter can still obtain state-of-the-art tracking performance if proper care is taken to fix the noise accumulated during occlusion. Instead of relying only on the linear state estimate (i.e., estimation-centric approach), we use object observations (i.e., the measurements by object detector) to compute a virtual trajectory over the occlusion period to fix the error accumulation of filter parameters. This allows more time steps to correct errors accumulated during occlusion. We name our method Observation-Centric SORT (OC-SORT). It remains Simple, Online, and Real-Time but improves robustness during occlusion and non-linear motion. Given off-the-shelf detections as input, OC-SORT runs at 700+ FPS on a single CPU. It achieves state-of-the-art on multiple datasets, including MOT17, MOT20, KITTI, head tracking, and especially DanceTrack where the object motion is highly non-linear. The code and models are available at https://github.com/noahcao/OC_SORT.

We tackle the task of Azimuth (angular dimension) super-resolution for Frequency Modulated Continuous Wave (FMCW) multiple-input multiple-output (MIMO) radar. FMCW MIMO radar is widely used in autonomous driving alongside Lidar and RGB cameras. However, compared to Lidar, MIMO radar is usually of low resolution due to hardware size restrictions. For example, achieving 1° azimuth resolution requires at least 100 receivers, but a single MIMO device usually supports at most 12 receivers. Having limitations on the number of receivers is problematic since a high-resolution measurement of azimuth angle is essential for estimating the location and velocity of objects. To improve the azimuth resolution of MIMO radar, we propose a light, yet efficient, Analog-to-Digital super-resolution model (ADC-SR) that predicts or hallucinates additional radar signals using signals from only a few receivers. Compared with the baseline models that are applied to processed radar Range-Azimuth-Doppler (RAD) maps, we show that our ADC-SR method that processes raw ADC signals achieves comparable performance with 98% (50 times) fewer parameters. We also propose a hybrid super-resolution model (Hybrid-SR) combining our ADC-SR with a standard RAD super-resolution model, and show that performance can be improved by a large margin. Experiments on our Pitt-Radar dataset and the RADIal dataset validate the importance of leveraging raw radar ADC signals. To assess the value of our super-resolution model for autonomous driving, we also perform object detection on the results of our super-resolution model and find that our super-resolution model improves detection performance by around 4% in mAP. The Pitt-Radar and the code will be released at the link.

We introduce a method for generating realistic pedestrian trajectories and full-body animations that can be controlled to meet user-defined goals. We draw on recent advances in guided diffusion modeling to achieve test-time controllability of trajectories, which is normally only associated with rule-based systems. Our guided diffusion model allows users to constrain trajectories through target waypoints, speed, and specified social groups while accounting for the surrounding environment context. This trajectory diffusion model is integrated with a novel physics-based humanoid controller to form a closed-loop, full-body pedestrian animation system capable of placing large crowds in a simulated environment with varying terrains. We further propose utilizing the value function learned during RL training of the animation controller to guide diffusion to produce trajectories better suited for particular scenarios such as collision avoidance and traversing uneven terrain. Video results are available on the project page.

We aim to improve the performance of the vehicle detection model with Lidar-Radar fusion and data augmentation. The recent works for Lidar-Radar fusion such as MVDNet or ST-MVDNet, have been proposed to have effective performance in detecting vehicles, and address the issue regarding missing modality. However, there are few works applying some global data augmentations such as rotation, translation, and scaling which are common for Lidar-only model. In order to further improve the previous Lidar-Radar fusion model, we propose a model named ST-MVDNet++ by leveraging the self-training teacher-student framework with integrating more common data augmentations such as global rotation, translation, and scaling. To ensure the data augmentations are consistent and matched across Lidar and Radar, we apply the augmentations on bird-eye-view coordinates. We also introduce the student-only augmentation for robust training of the student model with the consistency loss from teacher model. We demonstrate that our leveraging of global consistent Lidar-Radar augmentation improve the previous works by 1 ∼ 2% in all of the experimental settings.

We present a physics-based humanoid controller that achieves high-fidelity motion imitation and fault-tolerant behavior in the presence of noisy input (e.g. pose estimates from video or generated from language) and unexpected falls. Our controller scales up to learning ten thousand motion clips without using any external stabilizing forces and learns to naturally recover from fail-state. Given reference motion, our controller can perpetually control simulated avatars without requiring resets. At its core, we propose the progressive multiplicative control policy (PMCP), which dynamically allocates new network capacity to learn harder and harder motion sequences. PMCP allows efficient scaling for learning from large-scale motion databases and adding new tasks, such as fail-state recovery, without catastrophic forgetting. We demonstrate the effectiveness of our controller by using it to imitate noisy poses from video-based pose estimators and language-based motion generators in a live and real-time multi-person avatar use case.

Existing volumetric methods for predicting 3D human pose estimation are accurate, but computationally expensive and optimized for single time-step prediction. We present TEMPO, an efficient multi-view pose estimation model that learns a robust spatiotemporal representation, improving pose accuracy while also tracking and forecasting human pose. We significantly reduce computation compared to the state-of-the-art by recurrently computing per-person 2D pose features, fusing both spatial and temporal information into a single representation. In doing so, our model is able to use spatiotemporal context to predict more accurate human poses without sacrificing efficiency. We further use this representation to track human poses over time as well as predict future poses. Finally, we demonstrate that our model is able to generalize across datasets without scene-specific fine-tuning. TEMPO achieves 10% better MPJPE with a 33× improvement in FPS compared to TesseTrack on the challenging CMU Panoptic Studio dataset. Our code and demos are available at https://rccchoudhury.github.io/tempo2023/.

Accurately estimating 3D hand pose is crucial for understanding how humans interact with the world. Despite remarkable progress, existing methods often struggle to generate plausible hand poses when the hand is heavily occluded or blurred. In videos, the movements of the hand allow us to observe various parts of the hand that may be occluded or blurred in a single frame. To adaptively leverage the visual clue before and after the occlusion or blurring for robust hand pose estimation, we propose the Deformer: a framework that implicitly reasons about the relationship between hand parts within the same image (spatial dimension) and different timesteps (temporal dimension). We show that a naive application of the transformer self-attention mechanism is not sufficient because motion blur or occlusions in certain frames can lead to heavily distorted hand features and generate imprecise keys and queries. To address this challenge, we incorporate a Dynamic Fusion Module into Deformer, which predicts the deformation of the hand and warps the hand mesh predictions from nearby frames to explicitly support the current frame estimation. Furthermore, we have observed that errors are unevenly distributed across different hand parts, with vertices around fingertips having disproportionately higher errors than those around the palm. We mitigate this issue by introducing a new loss function called maxMSE that automatically adjusts the weight of every vertex to focus the model on critical hand parts. Extensive experiments show that our method significantly outperforms state-of-the-art methods by 10%, and is more robust to occlusions (over 14%).

We present EgoHumans, a new multi-view multi-human video benchmark to advance the state-of-the-art of egocentric human 3D pose estimation and tracking. Existing ego-centric benchmarks either capture single subject or indoor-only scenarios, which limit the generalization of computer vision algorithms for real-world applications. We propose a novel 3D capture setup to construct a comprehensive ego-centric multi-human benchmark in the wild with annotations to support diverse tasks such as human detection, tracking, 2D/3D pose estimation, and mesh recovery. We leverage consumer-grade wearable camera-equipped glasses for the egocentric view, which enables us to capture dynamic activities like playing tennis, fencing, volleyball, etc. Furthermore, our multi-view setup generates accurate 3D ground truth even under severe or complete occlusion. The dataset consists of more than 125k egocentric images, spanning diverse scenes with a particular focus on challenging and unchoreographed multi-human activities and fast-moving egocentric views. We rigorously evaluate existing state-of-the-art methods and highlight their limitations in the egocentric scenario, specifically on multi-human tracking. To address such limitations, we propose EgoFormer, a novel approach with a multi-stream transformer architecture and explicit 3D spatial reasoning to estimate and track the human pose. EgoFormer significantly outperforms prior art by 13.6% IDF1 on the EgoHumans dataset.

Multi-modal trajectory forecasting methods commonly evaluate using single-agent metrics (marginal metrics), such as minimum Average Displacement Error (ADE) and Final Displacement Error (FDE), which fail to capture joint performance of multiple interacting agents. Only focusing on marginal metrics can lead to unnatural predictions, such as colliding trajectories or diverging trajectories for people who are clearly walking together as a group. Consequently, methods optimized for marginal metrics lead to overly-optimistic estimations of performance, which is detrimental to progress in trajectory forecasting research. In response to the limitations of marginal metrics, we present the first comprehensive evaluation of state-of-the-art (SOTA) trajectory forecasting methods with respect to multi-agent metrics (joint metrics): JADE, JFDE, and collision rate. We demonstrate the importance of joint metrics as opposed to marginal metrics with quantitative evidence and qualitative examples drawn from the ETH / UCY and Stanford Drone datasets. We introduce a new loss function incorporating joint metrics that, when applied to a SOTA trajectory forecasting method, achieves a 7% improvement in JADE / JFDE on the ETH / UCY datasets with respect to the previous SOTA. Our results also indicate that optimizing for joint metrics naturally leads to an improvement in interaction modeling, as evidenced by a 16% decrease in mean collision rate on the ETH / UCY datasets with respect to the previous SOTA. Code is available at github.com/ericaweng/joint-metrics-matter.

Motion-based association for Multi-Object Tracking (MOT) has recently re-achieved prominence with the rise of powerful object detectors. Despite this, little work has been done to incorporate appearance cues beyond simple heuristic models that lack robustness to feature degradation. In this paper, we propose a novel way to leverage objects’ appearances to adaptively integrate appearance matching into existing high-performance motion-based methods. Building upon the pure motion-based method OC-SORT, we achieve 1st place on MOT20 and 2nd place on MOT17 with 63.9 and 64.9 HOTA, respectively. We also achieve 61.3 HOTA on the challenging DanceTrack benchmark as a new state-of-the-art even compared to more heavily-designed methods. The code and models are available at https://github.com/GerardMaggiolino/Deep-OC-SORT.

Considerable research effort has been devoted to LiDAR-based 3D object detection and empirical performance has been significantly improved. While progress has been en-couraging, we observe an overlooked issue: it is not yet common practice to compare different 3D detectors under the same cost, e.g., inference latency. This makes it difficult to quantify the true performance gain brought by recently proposed architecture designs. The goal of this work is to conduct a cost-aware evaluation of LiDAR-based 3D object detectors. Specifically, we focus on SECOND, a simple grid-based one-stage detector, and analyze its performance under different costs by scaling its original architecture. Then we compare the family of scaled SECOND with recent 3D detection methods, such as Voxel R-CNN and PV-RCNN++. The results are surprising. We find that, if allowed to use the same latency, SECOND can match the performance of PV-RCNN++, the current state-of-the-art method on the Waymo Open Dataset. Scaled SECOND also easily outperforms many recent 3D detection methods published during the past year. We recommend future research control the inference cost in their empirical comparison and include the family of scaled SECOND as a strong baseline when presenting novel 3D detection methods.

One of the recent trends in vision problems is to use natural language captions to describe the objects of interest. This approach can overcome some limitations of traditional methods that rely on bounding boxes or category annotations. This paper introduces a novel paradigm for Multiple Object Tracking called Type-to-Track, which allows users to track objects in videos by typing natural language descriptions. We present a new dataset for that Grounded Multiple Object Tracking task, called GroOT, that contains videos with various types of objects and their corresponding textual captions describing their appearance and action in detail. Additionally, we introduce two new evaluation protocols and formulate evaluation metrics specifically for this task. We develop a new efficient method that models a transformer-based eMbed-ENcoDE-extRact framework (MENDER) using the third-order tensor decomposition. The experiments in five scenarios show that our MENDER approach outperforms another two-stage design in terms of accuracy and efficiency, up to 14.7\% accuracy and $4\times$ speed faster.

We present an approach for 3D global human mesh recovery from monocular videos recorded with dynamic cameras. Our approach is robust to severe and long-term occlusions and tracks human bodies even when they go outside the camera's field of view. To achieve this, we first propose a deep generative motion infiller, which autoregressively infills the body motions of occluded humans based on visible motions. Additionally, in contrast to prior work, our approach reconstructs human meshes in consistent global coordinates even with dynamic cameras. Since the joint reconstruction of human motions and camera poses is underconstrained, we propose a global trajectory predictor that generates global human trajectories based on local body movements. Using the predicted trajectories as anchors, we present a global optimization framework that refines the predicted trajectories and optimizes the camera poses to match the video evidence such as 2D keypoints. Experiments on challenging indoor and in-the-wild datasets with dynamic cameras demonstrate that the proposed approach outperforms prior methods significantly in terms of motion infilling and global mesh recovery.

A key component of understanding hand-object interactions is the ability to identify the active object-the object that is being manipulated by the human hand. In order to accurately localize the active object, any method must reason using information encoded by each image pixel, such as whether it belongs to the hand, the object, or the background. To leverage each pixel as evidence to determine the bounding box of the active object, we propose a pixel-wise voting function. Our pixel-wise voting function takes an initial bounding box as input and produces an improved bounding box of the active object as output. The voting function is designed so that each pixel inside of the input bounding box votes for an improved bounding box, and the box with the majority vote is selected as the output. We call the collection of bounding boxes generated inside of the voting function, the Relational Box Field, as it characterizes a field of bounding boxes defined in relationship to the current bounding box. While our voting function is able to improve the bounding box of the active object, one round of voting is typically not enough to accurately localize the active object. Therefore, we repeatedly apply the voting function to sequentially improve the location of the bounding box. However, since it is known that repeatedly applying a one-step predictor (i.e., auto-regressive processing with our voting function) can cause a data distribution shift, we mitigate this issue using reinforcement learning (RL). We adopt standard RL to learn the voting function parameters and show that it provides a meaningful improvement over a standard supervised learning approach. We perform experiments on two large-scale datasets: 100DOH and MECCANO, improving AP50 performance by 8% and 30%, respectively, over the state of the art. The project page with code and visualizations can be found at https://fuqichen1998.github.io/SequentialVotingDet/.

We introduce Ego4D, a massive-scale egocentric video dataset and benchmark suite. It offers 3,670 hours of dailylife activity video spanning hundreds of scenarios (household, outdoor, workplace, leisure, etc.) captured by 931 unique camera wearers from 74 worldwide locations and 9 different countries. The approach to collection is designed to uphold rigorous privacy and ethics standards, with consenting participants and robust de-identification procedures where relevant. Ego4D dramatically expands the volume of diverse egocentric video footage publicly available to the research community. Portions of the video are accompanied by audio, 3D meshes of the environment, eye gaze, stereo, and/or synchronized videos from multiple egocentric cameras at the same event. Furthermore, we present a host of new benchmark challenges centered around understanding the first-person visual experience in the past (querying an episodic memory), present (analyzing hand-object manipulation, audio-visual conversation, and social interactions), and future (forecasting activities). By publicly sharing this massive annotated dataset and benchmark suite, we aim to push the frontier of first-person perception. Project page: https://ego4d-data.org/

Top-down methods for monocular human mesh recovery have two stages: (1) detect human bounding boxes; (2) treat each bounding box as an independent single-human mesh recovery task. Unfortunately, the single-human assumption does not hold in images with multi-human occlusion and crowding. Consequently, top-down methods have difficulties in recovering accurate 3D human meshes under severe person-person occlusion. To address this, we present Occluded Human Mesh Recovery (OCHMR) - a novel top-down mesh recovery approach that incorporates image spatial context to overcome the limitations of the single-human assumption. The approach is conceptually simple and can be applied to any existing top-down architecture. Along with the input image, we condition the top-down model on spatial context from the image in the form of body-center heatmaps. To reason from the predicted body centermaps, we introduce Contextual Normalization (CoNorm) blocks to adaptively modulate intermediate features of the top-down model. The contextual conditioning helps our model disambiguate between two severely overlapping human boundingboxes, making it robust to multi-person occlusion. Compared with state-of-the-art methods, OCHMR achieves superior performance on challenging multi-person benchmarks like 3DPW, CrowdPose and OCHuman. Specifically, our proposed contextual reasoning architecture applied to the SPIN model with ResNet-50 backbone results in 75.2 PMPJPE on 3DPW-PC, 23.6 AP on CrowdPose and 37.7 AP on OCHu- man datasets, a significant improvement of 6.9 mm, 6.4 AP and 20.8 AP respectively over the baseline.

We address the task of domain adaptation in object detection, where there is an obvious domain gap between a domain with annotations (source) and a domain of interest without annotations (target). As a popular semi-supervised learning method, the teacher-student framework (a student model is supervised by the pseudo labels from a teacher model) has also yielded a large accuracy gain in cross-domain object detection. However, it suffers from the domain shift and generates many low-quality pseudo labels (e.g., false positives), which leads to sub-optimal performance. To mitigate this problem, we propose a teacher-student framework named Adaptive Teacher (AT) which leverages domain adversarial learning and weak-strong data augmentation to address the domain gap. Specifically, we employ feature-level adversarial training in the student model, allowing features derived from the source and target domains to share similar distributions. This process ensures the student model produces domain-invariant features. Furthermore, we apply weak-strong augmentation and mutual learning between the teacher model (taking data from the target domain) and the student model (taking data from both domains). This enables the teacher model to learn the knowledge from the student model without being biased to the source domain. We show that AT demonstrates superiority over existing approaches and even Oracle (fully-supervised) models by a large margin. For example, we achieve 50.9% (49.3%) mAP on Foggy Cityscape (Cli-part1K), which is 9.2% (5.2%) and 8.2% (11.0%) higher than previous state-of-the-art and Oracle, respectively.

Fusion of multiple sensor modalities such as camera, Lidar, and Radar, which are commonly found on autonomous vehicles, not only allows for accurate detection but also robustifies perception against adverse weather conditions and individual sensor failures. Due to inherent sensor characteristics, Radar performs well under extreme weather conditions (snow, rain, fog) that significantly degrade camera and Lidar. Recently, a few works have developed vehicle detection methods fusing Lidar and Radar signals, i.e., MVD-Net. However, these models are typically developed under the assumption that the models always have access to two error-free sensor streams. If one of the sensors is unavailable or missing, the model may fail catastrophically. To mitigate this problem, we propose the Self-Training Multimodal Vehicle Detection Network (ST-MVDNet) which leverages a Teacher-Student mutual learning framework and a simulated sensor noise model used in strong data augmentation for Lidar and Radar. We show that by (1) enforcing output consistency between a Teacher network and a Student network and by (2) introducing missing modalities (strong augmentations) during training, our learned model breaks away from the error-free sensor assumption. This consistency enforcement enables the Student model to handle missing data properly and improve the Teacher model by updating it with the Student model's exponential moving average. Our experiments demonstrate that our proposed learning framework for multi-modal detection is able to better handle missing sensor data during inference. Furthermore, our method achieves new state-of-the-art performance (5% gain) on the Oxford Radar Robotcar dataset under various evaluation settings.

A typical pipeline for multi-object tracking (MOT) is to use a detector for object localization, and following re-identification (re-ID)for object association. This pipeline is partially motivated by recent progress in both object detection and re- ID, and partially motivated by biases in existing tracking datasets, where most objects tend to have distin-guishing appearance and re-ID models are sufficient for es-tablishing associations. In response to such bias, we would like to re-emphasize that methods for multi-object tracking should also work when object appearance is not sufficiently discriminative. To this end, we propose a large-scale dataset for multi-human tracking, where humans have sim-ilar appearance, diverse motion and extreme articulation. As the dataset contains mostly group dancing videos, we name it “DanceTrack”. We expect DanceTrack to provide a better platform to develop more MOT algorithms that rely less on visual discrimination and depend more on motion analysis. We benchmark several state-of-the-art trackers on our dataset and observe a significant performance drop on DanceTrack when compared against existing benchmarks. The dataset, project code and competition is released at: https://github.com/DanceTrack.

Multi-agent trajectory prediction is critical for planning and decision-making in human-interactive autonomous systems, such as self-driving cars. However, most prediction models are developed separately from their upstream perception (detection and tracking) modules, assuming ground truth past trajectories as inputs. As a result, their performance degrades significantly when using real-world noisy tracking results as inputs. This is typically caused by the propagation of errors from tracking to prediction, such as noisy tracks, fragments and identity switches. To alleviate this propagation of errors, we propose a new prediction paradigm that uses detections and their affinity matrices across frames as inputs, removing the need for error- prone data association during tracking. Since affinity matrices contain “soft” information about the similarity and identity of detections across frames, making prediction directly from affinity matrices retains strictly more information than making prediction from the tracklets generated by data association. Experiments on large-scale, real-world autonomous driving datasets show that our affinity-based prediction scheme11Our project website is at https://www.xinshuoweng.com/projects/Affinipred. reduces overall prediction errors by up to 57.9%, in comparison to standard prediction pipelines that use tracklets as inputs, with even more significant error reduction (up to 88.6%) if restricting the evaluation to challenging scenarios with tracking errors.

A popular paradigm in robotic learning is to train a policy from scratch for every new robot. This is not only inefficient but also often impractical for complex robots. In this work, we consider the problem of transferring a policy across two different robots with significantly different parameters such as kinematics and morphology. Existing approaches that train a new policy by matching the action or state transition distribution, including imitation learning methods, fail due to optimal action and/or state distribution being mismatched in different robots. In this paper, we propose a novel method named REvolveR of using continuous evolutionary models for robotic policy transfer implemented in a physics simulator. We interpolate between the source robot and the target robot by finding a continuous evolutionary change of robot parameters. An expert policy on the source robot is transferred through training on a sequence of intermediate robots that gradually evolve into the target robot. Experiments on a physics simulator show that the proposed continuous evolutionary model can effectively transfer the policy across robots and achieve superior sample efficiency on new robots. The proposed method is especially advantageous in sparse reward settings where exploration can be significantly reduced.

Indoor localization systems often fuse inertial odometry with map information via hand-defined methods to reduce odometry drift, but such methods are sensitive to noise and struggle to generalize across odometry sources. To address the robustness problem in map utilization, we propose a data-driven prior on possible user locations in a map by combining learned spatial map embeddings and temporal odometry embeddings. Our prior learns to encode which map regions are feasible locations for a user more accurately than previous hand-defined methods. This prior leads to a 49% improvement in inertial-only localization accuracy when used in a particle filter. This result is significant, as it shows that our relative positioning method can match the performance of absolute positioning using bluetooth beacons. To show the gen-eralizability of our method, we also show similar improvements using wheel encoder odometry. Our code will be made publicly available†1project page: https://rebrand.ly/learned-map-prior.

We propose embodied scene-aware human pose estimation where we estimate 3D poses based on a simulated agent's proprioception and scene awareness, along with external third-person observations. Unlike prior methods that often resort to multistage optimization, non-causal inference, and complex contact modeling to estimate human pose and human scene interactions, our method is one-stage, causal, and recovers global 3D human poses in a simulated environment. Since 2D third-person observations are coupled with the camera pose, we propose to disentangle the camera pose and use a multi-step projection gradient defined in the global coordinate frame as the movement cue for our embodied agent. Leveraging a physics simulation and prescanned scenes (e.g., 3D mesh), we simulate our agent in everyday environments (library, office, bedroom, etc.) and equip our agent with environmental sensors to intelligently navigate and interact with the geometries of the scene. Our method also relies only on 2D keypoints and can be trained on synthetic datasets derived from popular human motion databases. To evaluate, we use the popular H36M and PROX datasets and achieve high quality pose estimation on the challenging PROX dataset without ever using PROX motion sequences for training. Code and videos are available on the project page.

The goal of No-Reference Image Quality Assessment (NR-IQA) is to estimate the perceptual image quality in accordance with subjective evaluations, it is a complex and unsolved problem due to the absence of the pristine reference image. In this paper, we propose a novel model to address the NR-IQA task by leveraging a hybrid approach that benefits from Convolutional Neural Networks (CNNs) and self-attention mechanism in Transformers to extract both local and non-local features from the input image. We capture local structure information of the image via CNNs, then to circumvent the locality bias among the extracted CNNs features and obtain a non-local representation of the image, we utilize Transformers on the extracted features where we model them as a sequential input to the Transformer model. Furthermore, to improve the monotonicity correlation between the subjective and objective scores, we utilize the relative distance information among the images within each batch and enforce the relative ranking among them. Last but not least, we observe that the performance of NR-IQA models degrades when we apply equivariant transformations (e.g. horizontal flipping) to the inputs. Therefore, we propose a method that leverages self-consistency as a source of self-supervision to improve the robustness of NR-IQA models. Specifically, we enforce self-consistency between the outputs of our quality assessment model for each image and its transformation (horizontally flipped) to utilize the rich self-supervisory information and reduce the uncertainty of the model. To demonstrate the effectiveness of our work, we evaluate it on seven standard IQA datasets (both synthetic and authentic) and show that our model achieves state-of-the-art results on various datasets.1

Various methods for solving the inverse reinforcement learning (IRL) problem have been developed independently in machine learning and economics. In particular, the method of Maximum Causal Entropy IRL is based on the perspective of entropy maximization, while related advances in the field of economics instead assume the existence of unobserved action shocks to explain expert behavior (Nested Fixed Point Algorithm, Conditional Choice Probability method, Nested Pseudo-Likelihood Algorithm). In this work, we make previously unknown connections between these related methods from both fields. We achieve this by showing that they all belong to a class of optimization problems, characterized by a common form of the objective, the associated policy and the objective gradient. We demonstrate key computational and algorithmic differences which arise between the methods due to an approximation of the optimal soft value function, and describe how this leads to more efficient algorithms. Using insights which emerge from our study of this class of optimization problems, we identify various problem scenarios and investigate each method's suitability for these problems.

Many smartphone applications use inertial measurement units (IMUs) to sense movement, but the use of these sensors for pedestrian localization can be challenging due to their noise characteristics. Recent data-driven inertial odometry approaches have demonstrated the increasing feasibility of inertial navigation. However, they still rely upon conventional smartphone orientation estimates that they assume to be accurate, while in fact these orientation estimates can be a significant source of error. To address the problem of inaccurate orientation estimates, we present a two-stage, data-driven pipeline using a commodity smartphone that first estimates device orientations and then estimates device position. The orientation module relies on a recurrent neural network and Extended Kalman Filter to obtain orientation estimates that are used to then rotate raw IMU measurements into the appropriate reference frame. The position module then passes those measurements through another recurrent network architecture to perform localization. Our proposed method outperforms state-of-the-art methods in both orientation and position error on a large dataset we constructed that contains 20 hours of pedestrian motion across 3 buildings and 15 subjects. Code and data are available at https://github.com/KlabCMU/IDOL.

Accurate estimation of 3D human motion from monocular video requires modeling both kinematics (body motion without physical forces) and dynamics (motion with physical forces). To demonstrate this, we present SimPoE, a Simulation-based approach for 3D human Pose Estimation, which integrates image-based kinematic inference and physics-based dynamics modeling. SimPoE learns a policy that takes as input the current-frame pose estimate and the next image frame to control a physically-simulated character to output the next-frame pose estimate. The policy contains a learnable kinematic pose refinement unit that uses 2D keypoints to iteratively refine its kinematic pose estimate of the next frame. Based on this refined kinematic pose, the policy learns to compute dynamics-based control (e.g., joint torques) of the character to advance the current-frame pose estimate to the pose estimate of the next frame. This design couples the kinematic pose refinement unit with the dynamics-based control generation unit, which are learned jointly with reinforcement learning to achieve accurate and physically-plausible pose estimation. Furthermore, we propose a meta-control mechanism that dynamically adjusts the character’s dynamics parameters based on the character state to attain more accurate pose estimates. Experiments on large-scale motion datasets demonstrate that our approach establishes the new state of the art in pose accuracy while ensuring physical plausibility.

We address the problem of estimating the 3D pose of a network of cameras for large-environment wide-baseline scenarios, e.g., cameras for construction sites, sports stadiums, and public spaces. This task is challenging since detecting and matching the same 3D keypoint observed from two very different camera views is difficult, making standard structure-from-motion (SfM) pipelines inapplicable. In such circumstances, treating people in the scene as "keypoints" and associating them across different camera views can be an alternative method for obtaining correspondences. Based on this intuition, we propose a method that uses ideas from person re-identification (re-ID) for wide-baseline camera calibration. Our method first employs a re-ID method to associate human bounding boxes across cameras, then converts bounding box correspondences to point correspondences, and finally solves for camera pose using multi-view geometry and bundle adjustment. Since our method does not require specialized calibration targets except for visible people, it applies to situations where frequent calibration updates are required. We perform extensive experiments on datasets captured from scenes of different sizes (80m2, 350m2, 600m2), camera settings (indoor and outdoor), and human activities (walking, playing basketball, construction). Experiment results show that our method achieves similar performance to standard SfM methods relying on manually labeled point correspondences.

Predicting accurate future trajectories of multiple agents is essential for autonomous systems but is challenging due to the complex interaction between agents and the uncertainty in each agent’s future behavior. Forecasting multi-agent trajectories requires modeling two key dimensions: (1) time dimension, where we model the influence of past agent states over future states; (2) social dimension, where we model how the state of each agent affects others. Most prior methods model these two dimensions separately, e.g., first using a temporal model to summarize features over time for each agent independently and then modeling the interaction of the summarized features with a social model. This approach is suboptimal since independent feature encoding over either the time or social dimension can result in a loss of information. Instead, we would prefer a method that allows an agent’s state at one time to directly affect another agent’s state at a future time. To this end, we propose a new Transformer, termed AgentFormer, that simultaneously models the time and social dimensions. The model leverages a sequence representation of multi-agent trajectories by flattening trajectory features across time and agents. Since standard attention operations disregard the agent identity of each element in the sequence, AgentFormer uses a novel agent-aware attention mechanism that preserves agent identities by attending to elements of the same agent differently than elements of other agents. Based on AgentFormer, we propose a stochastic multi-agent trajectory prediction model that can attend to features of any agent at any previous timestep when inferring an agent’s future position. The latent intent of all agents is also jointly modeled, allowing the stochasticity in one agent’s behavior to affect other agents. Extensive experiments show that our method significantly improves the state of the art on well-established pedestrian and autonomous driving datasets.

We present RePOSE, a fast iterative refinement method for 6D object pose estimation. Prior methods perform refinement by feeding zoomed-in input and rendered RGB images into a CNN and directly regressing an update of a refined pose. Their runtime is slow due to the computational cost of CNN, which is especially prominent in multiple-object pose refinement. To overcome this problem, RePOSE leverages image rendering for fast feature extraction using a 3D model with a learnable texture. We call this deep texture rendering, which uses a shallow multilayer perceptron to directly regress a view-invariant image representation of an object. Furthermore, we utilize differentiable Levenberg-Marquardt (LM) optimization to refine a pose fast and accurately by minimizing the distance between the input and rendered image representations without the need of zooming in. These image representations are trained such that differentiable LM optimization converges within few iterations. Consequently, RePOSE runs at 92 FPS and achieves state-of-the-art accuracy of 51.6% on the Occlusion LineMOD dataset - a 4.1% absolute improvement over the prior art, and comparable result on the YCB-Video dataset with a much faster runtime. The code is available at https://github.com/sh8/repose.

We address the task of Re-Identification (Re-ID) in multi-target multi-camera (MTMC) tracking where we track multiple pedestrians using multiple overlapping uncalibrated (unknown pose) cameras. Since the videos are temporally synchronized and spatially overlapping, we can see a person from multiple views and associate their trajectory across cameras. In order to find the correct association between pedestrians visible from multiple views during the same time window, we extract a visual feature from a tracklet (sequence of pedestrian images) that encodes its similarity and dissimilarity to all other candidate tracklets. We propose a inter-tracklet (person to person) attention mechanism that learns a representation for a target tracklet while taking into account other tracklets across multiple views. Furthermore, to encode the gait and motion of a person, we introduce second intra-tracklet (person-specific) attention module with position embeddings. This second module employs a transformer encoder to learn a feature from a sequence of features over one tracklet. Experimental results on WILDTRACK and our new dataset ‘ConstructSite’ confirm the superiority of our model over state-of-the-art ReID methods (5% and 10% performance gain respectively) in the context of uncalibrated MTMC tracking. While our model is designed for overlapping cameras, we also obtain state-of-the-art results on two other benchmark datasets (MARS and DukeMTMC) with non-overlapping cameras.

We present a large-scale stereo RGB image object pose estimation dataset named the StereOBJ-1M dataset. The dataset is designed to address challenging cases such as object transparency, translucency, and specular reflection, in addition to the common challenges of occlusion, symmetry, and variations in illumination and environments. In order to collect data of sufficient scale for modern deep learning models, we propose a novel method for efficiently annotating pose data in a multi-view fashion that allows data capturing in complex and flexible environments. Fully annotated with 6D object poses, our dataset contains over 396K frames and over 1.5M annotations of 18 objects recorded in 183 scenes constructed in 11 different environments. The 18 objects include 8 symmetric objects, 7 transparent objects, and 8 reflective objects. We benchmark two state-of-the-art pose estimation frameworks on StereOBJ-1M as baselines for future work. We also propose a novel object-level pose optimization method for computing 6D pose from keypoint predictions in multiple images.

LiDAR sensors can be used to obtain a wide range of measurement signals other than a simple 3D point cloud, and those signals can be leveraged to improve perception tasks like 3D object detection. A single laser pulse can be partially reflected by multiple objects along its path, resulting in multiple measurements called echoes. Multi-echo measurement can provide information about object contours and semi-transparent surfaces which can be used to better identify and locate objects. LiDAR can also measure surface reflectance (intensity of laser pulse return), as well as ambient light of the scene (sunlight reflected by objects). These signals are already available in commercial LiDAR devices but have not been used in most LiDAR-based detection models. We present a 3D object detection model which leverages the full spectrum of measurement signals provided by LiDAR. First, we propose a multi-signal fusion (MSF) module to combine (1) the reflectance and ambient features extracted with a 2D CNN, and (2) point cloud features extracted using a 3D graph neural network (GNN). Second, we propose a multi-echo aggregation (MEA) module to combine the information encoded in different sets of echo points. Compared with traditional single echo point cloud methods, our proposed Multi-Signal LiDAR Detector (MSLiD) extracts richer context information from a wider range of sensing measurements and achieves more accurate 3D object detection. Experiments show that by incorporating the multi-modality of LiDAR, our method outperforms the state-of-the-art by up to relatively 9.1%.

DETR is a recently proposed Transformer-based method which views object detection as a set prediction problem and achieves state-of-the-art performance but demands extra-long training time to converge. In this paper, we investigate the causes of the optimization difficulty in the training of DETR. Our examinations reveal several factors contributing to the slow convergence of DETR, primarily the issues with the Hungarian loss and the Transformer cross-attention mechanism. To overcome these issues we propose two solutions, namely, TSP-FCOS (Transformer-based Set Prediction with FCOS) and TSP-RCNN (Transformer-based Set Prediction with RCNN). Experimental results show that the proposed methods not only converge much faster than the original DETR, but also significantly outperform DETR and other baselines in terms of detection accuracy. Code is released at https://github.com/Edward-Sun/TSP-Detection.

Detecting small cracks in concrete is difficult due to the complexity and thinness of cracking patterns, which requires the development of refined vision-based segmentation algorithms that can accurately characterize the details of crack defects. While existing methods are good at generally outlining cracks, due to inherent differences in shape distributions between common objects and cracks, their predictions often have disconnected segments and inaccuracy along boundaries. To this end, we develop a refinement framework using reinforcement learning (RL) that can better recognize details specific to cracks. Our method uses an RL agent to iteratively improve per-pixel crack predictions of a general segmentation model. We find that in addition to connecting gaps in predictions, the RL agent is also able to detect cracks that are missed in the original predictions. It does so by using the originally detected regions as crack priors to branch out from. Refining outputs of a commonly used per-pixel segmentation model, our method outperforms the current state-of-the-art approaches for crack segmentation. Our experiments also demonstrate that our method generalizes well to a similar task of vessel segmentation.

Object detection and data association are critical components in multi-object tracking (MOT) systems. Despite the fact that the two components are dependent on each other, prior works often design detection and data association modules separately which are trained with separate objectives. As a result, one cannot back-propagate the gradients and optimize the entire MOT system, which leads to sub-optimal performance. To address this issue, recent works simultaneously optimize detection and data association modules under a joint MOT framework, which has shown improved performance in both modules. In this work, we propose a new instance of joint MOT approach based on Graph Neural Networks (GNNs). The key idea is that GNNs can model relations between variablesized objects in both the spatial and temporal domains, which is essential for learning discriminative features for detection and data association. Through extensive experiments on the MOT15/16/17/20 datasets, we demonstrate the effectiveness of our GNN-based joint MOT approach and show state-of-the-art performance for both detection and MOT tasks.

We present KDFNet, a novel method for 6D object pose estimation from RGB images. To handle occlusion, many recent works have proposed to localize 2D keypoints through pixel-wise voting and solve a Perspective-n-Point (PnP) problem for pose estimation, which achieves leading performance. However, such voting process is direction-based and cannot handle long and thin objects where the direction intersections cannot be robustly found. To address this problem, we propose a novel continuous representation called Keypoint Distance Field (KDF) for projected 2D keypoint locations. Formulated as a 2D array, each element of the KDF stores the 2D Euclidean distance between the corresponding image pixel and a specified projected 2D keypoint. We use a fully convolutional neural network to regress the KDF for each keypoint. Using this KDF encoding of projected object keypoint locations, we propose to use a distance-based voting scheme to localize the keypoints by calculating circle intersections in a RANSAC fashion. We validate the design choices of our framework by extensive ablation experiments. Our proposed method achieves state-of-the-art performance on Occlusion LINEMOD dataset with an average ADD(-S) accuracy of 50.3% and TOD dataset mug subset with an average ADD accuracy of 75.72%. Extensive experiments and visualizations demonstrate that the proposed method is able to robustly estimate the 6D pose in challenging scenarios including occlusion.

We propose a method for object-aware 3D egocentric pose estimation that tightly integrates kinematics modeling, dynamics modeling, and scene object information. Unlike prior kinematics or dynamics-based approaches where the two components are used disjointly, we synergize the two approaches via dynamics-regulated training. At each timestep, a kinematic model is used to provide a target pose using video evidence and simulation state. Then, a prelearned dynamics model attempts to mimic the kinematic pose in a physics simulator. By comparing the pose instructed by the kinematic model against the pose generated by the dynamics model, we can use their misalignment to further improve the kinematic model. By factoring in the 6DoF pose of objects (e.g., chairs, boxes) in the scene, we demonstrate for the first time, the ability to estimate physically-plausible 3D human-object interactions using a single wearable camera. We evaluate our egocentric pose estimation method in both controlled laboratory settings and real-world scenarios.

Person re-identification (re-ID) aims to recognize instances of the same person contained in multiple images taken across different cameras. Existing methods for re-ID tend to rely heavily on the assumption that both query and gallery images of the same person have the same clothing. Unfortunately, this assumption may not hold for datasets captured over long periods of time. To tackle the re-ID problem in the context of clothing changes, we propose a novel representation learning method which is able to generate a shape-based feature representation that is invariant to clothing. We call our model the Clothing Agnostic Shape Extraction Network (CASE-Net). CASE-Net learns a representation of a person that depends primarily on shape via adversarial learning and feature disentanglement. Quantitative and qualitative results across 5 datasets (Div-Market, Market1501, three large-scale datasets under clothing changes) show our approach makes significant improvements over prior state-of-the-art approaches.

Social media platforms feature short animations known as GIFs, but they are inaccessible to people with vision impairments. Unlike static images, GIFs contain action and visual indications of sound, which can be challenging to describe in alternative text descriptions. We examine a large sample of inaccessible GIFs on Twitter to document how they are used and what visual elements they contain. In interviews with 10 blind Twitter users, we discuss what elements of GIF content should be described and their experiences with GIFs online. The participants compared alternative text descriptions with two other alternative audio formats: (i) the original audio from the GIF source video and (ii) a spoken audio description. We recommend that social media platforms automatically include alt text descriptions for popular GIFs (as Twitter has begun to do), and content producers create audio descriptions to ensure everyone has a rich and emotive experience with GIFs online.

We present ReCog, a mobile app that enables blind users to recognize objects by training a deep network with their own photos of such objects. This functionality is useful to differentiate personal objects, which cannot be recognized with pre-trained recognizers and may lack distinguishing tactile features. To ensure that the objects are well-framed in the captured photos, ReCog integrates a camera-aiming guidance that tracks target objects and instructs the user through verbal and sonification feedback to appropriately frame them.

Social media platforms are integral to public and private discourse, but are becoming less accessible to people with vision impairments due to an increase in user-posted images. Some platforms (i.e. Twitter) let users add image descriptions (alternative text), but only 0.1% of images include these. To address this accessibility barrier, we created Twitter A11y, a browser extension to add alternative text on Twitter using six methods. For example, screenshots of text are common, so we detect textual images, and create alternative text using optical character recognition. Twitter A11y also leverages services to automatically generate alternative text or reuse them from across the web. We compare the coverage and quality of Twitter A11y's six alt-text strategies by evaluating the timelines of 50 self-identified blind Twitter users. We find that Twitter A11y increases alt-text coverage from 7.6% to 78.5%, before crowdsourcing descriptions for the remaining images. We estimate that 57.5% of returned descriptions are high-quality. We then report on the experiences of 10 participants with visual impairments using the tool during a week-long deployment. Twitter A11y increases access to social media platforms for people with visual impairments by providing high-quality automatic descriptions for user-posted images.

Automatically reasoning about future human behaviors is a difficult problem but has significant practical applications to assistive systems. Part of this difficulty stems from learning systems' inability to represent all kinds of behaviors. Some behaviors, such as motion, are best described with continuous representations, whereas others, such as picking up a cup, are best described with discrete representations. Furthermore, human behavior is generally not fixed: people can change their habits and routines. This suggests these systems must be able to learn and adapt continuously. In this work, we develop an efficient deep generative model to jointly forecast a person's future discrete actions and continuous motions. On a large-scale egocentric dataset, EPIC-KITCHENS, we observe our method generates high-quality and diverse samples while exhibiting better generalization than related generative models. Finally, we propose a variant to continually learn our model from streaming data, observe its practical effectiveness, and theoretically justify its learning efficiency.

We describe a method for 3D human pose estimation from transient images (i.e., a 3D spatio-temporal histogram of photons) acquired by an optical non-line-of-sight (NLOS) imaging system. Our method can perceive 3D human pose by 'looking around corners' through the use of light indirectly reflected by the environment. We bring together a diverse set of technologies from NLOS imaging, human pose estimation and deep reinforcement learning to construct an end-to-end data processing pipeline that converts a raw stream of photon measurements into a full 3D human pose sequence estimate. Our contributions are the design of data representation process which includes (1) a learnable inverse point spread function (PSF) to convert raw transient images into a deep feature vector; (2) a neural humanoid control policy conditioned on the transient image feature and learned from interactions with a physics simulator; and (3) a data synthesis and augmentation strategy based on depth data that can be transferred to a real-world NLOS imaging system. Our preliminary experiments suggest that our method is able to generalize to real-world NLOS measurement to estimate physically-valid 3D human poses.

3D Multi-object tracking (MOT) is crucial to autonomous systems. Recent work uses a standard tracking-by-detection pipeline, where feature extraction is first performed independently for each object in order to compute an affinity matrix. Then the affinity matrix is passed to the Hungarian algorithm for data association. A key process of this standard pipeline is to learn discriminative features for different objects in order to reduce confusion during data association. In this work, we propose two techniques to improve the discriminative feature learning for MOT: (1) instead of obtaining features for each object independently, we propose a novel feature interaction mechanism by introducing the Graph Neural Network. As a result, the feature of one object is informed of the features of other objects so that the object feature can lean towards the object with similar feature (i.e., object probably with a same ID) and deviate from objects with dissimilar features (i.e., object probably with different IDs), leading to a more discriminative feature for each object; (2) instead of obtaining the feature from either 2D or 3D space in prior work, we propose a novel joint feature extractor to learn appearance and motion features from 2D and 3D space simultaneously. As features from different modalities often have complementary information, the joint feature can be more discriminate than feature from each individual modality. To ensure that the joint feature extractor does not heavily rely on one modality, we also propose an ensemble training paradigm. Through extensive evaluation, our proposed method achieves state-of-the-art performance on KITTI and nuScenes 3D MOT benchmarks. Our code will be made available at https://github.com/xinshuoweng/GNN3DMOT.

We aim to enable robots to visually localize a target person through the aid of an additional sensing modality - the target person's 3D inertial measurements. The need for such technology may arise when a robot is to meet a person in a crowd for the first time or when an autonomous vehicle must rendezvous with a rider amongst a crowd without knowing the appearance of the person in advance. A person's inertial information can be measured with a wearable device such as a smart-phone and can be shared selectively with an autonomous system during the rendezvous. We propose a method to learn a visual-inertial feature space in which the motion of a person in video can be easily matched to the motion measured by a wearable inertial measurement unit (IMU). The transformation of the two modalities into the joint feature space is learned through the use of a triplet loss which forces inertial motion features and video motion features generated by the same person to lie close in the joint feature space. To validate our approach, we compose a dataset of over 3,000 video segments of moving people along with wearable IMU data. We show that our method is able to localize a target person with 80.7% accuracy averaged over testing data with various number of candidates using only 5 seconds of IMU data and video.

3D multi-object tracking (MOT) is an essential component for many applications such as autonomous driving and assistive robotics. Recent work on 3D MOT focuses on developing accurate systems giving less attention to practical considerations such as computational cost and system complexity. In contrast, this work proposes a simple real-time 3D MOT system. Our system first obtains 3D detections from a LiDAR point cloud. Then, a straightforward combination of a 3D Kalman filter and the Hungarian algorithm is used for state estimation and data association. Additionally, 3D MOT datasets such as KITTI evaluate MOT methods in the 2D space and standardized 3D MOT evaluation tools are missing for a fair comparison of 3D MOT methods. Therefore, we propose a new 3D MOT evaluation tool along with three new metrics to comprehensively evaluate 3D MOT methods. We show that, although our system employs a combination of classical MOT modules, we achieve state-of-the-art 3D MOT performance on two 3D MOT benchmarks (KITTI and nuScenes). Surprisingly, although our system does not use any 2D data as inputs, we achieve competitive performance on the KITTI 2D MOT leaderboard. Our proposed system runs at a rate of 207.4 FPS on the KITTI dataset, achieving the fastest speed among all modern MOT systems. To encourage standardized 3D MOT evaluation, our code is publicly available at http://www.xinshuoweng.com/projects/AB3DMOT.

Reinforcement learning has shown great promise for synthesizing realistic human behaviors by learning humanoid control policies from motion capture data. However, it is still very challenging to reproduce sophisticated human skills like ballet dance, or to stably imitate long-term human behaviors with complex transitions. The main difficulty lies in the dynamics mismatch between the humanoid model and real humans. That is, motions of real humans may not be physically possible for the humanoid model. To overcome the dynamics mismatch, we propose a novel approach, residual force control (RFC), that augments a humanoid control policy by adding external residual forces into the action space. During training, the RFC-based policy learns to apply residual forces to the humanoid to compensate for the dynamics mismatch and better imitate the reference motion. Experiments on a wide range of dynamic motions demonstrate that our approach outperforms state-of-the-art methods in terms of convergence speed and the quality of learned motions. Notably, we showcase a physics-based virtual character empowered by RFC that can perform highly agile ballet dance moves such as pirouette, arabesque and jeté. Furthermore, we propose a dual-policy control framework, where a kinematic policy and an RFC-based policy work in tandem to synthesize multi-modal infinite-horizon human motions without any task guidance or user input. Our approach is the first humanoid control method that successfully learns from a large-scale human motion dataset (Human3.6M) and generates diverse long-term motions. Code and videos are available at https://www.ye-yuan.com/rfc.

We present MonoEye, a multimodal human motion capture system using a single RGB camera with an ultra-wide fisheye lens, mounted on the user's chest. Existing optical motion capture systems use multiple cameras, which are synchronized and require camera calibration. These systems also have usability constraints that limit the user's movement and operating space. Since the MonoEye system is based on a wearable single RGB camera, the wearer's 3D body pose can be captured without space and environment limitations. The body pose, captured with our system, is aware of the camera orientation and therefore it is possible to recognize various motions that existing egocentric motion capture systems cannot recognize. Furthermore, the proposed system captures not only the wearer's body motion but also their viewport using the head pose estimation and an ultra-wide image. To implement robust multimodal motion capture, we design three deep neural networks: BodyPoseNet, HeadPoseNet, and CameraPoseNet, that estimate 3D body pose, head pose, and camera pose in real-time, respectively. We train these networks with our new extensive synthetic dataset providing 680K frames of renderings of people with a wide range of body shapes, clothing, actions, backgrounds, and lighting conditions. To demonstrate the interactive potential of the MonoEye system, we present several application examples from common body gestural to context-aware interactions.

The automatic recognition of how people use their hands and fingers in natural settings -- without instrumenting the fingers -- can be useful for many mobile computing applications. To achieve such an interface, we propose a vision-based 3D hand pose estimation framework using a wrist-worn camera. The main challenge is the oblique angle of the wrist-worn camera, which makes the fingers scarcely visible. To address this, a special network that observes deformations on the back of the hand is required. We introduce DorsalNet, a two-stream convolutional neural network to regress finger joint angles from spatio-temporal features of the dorsal hand region (the movement of bones, muscle, and tendons). This work is the first vision-based real-time 3D hand pose estimator using visual features from the dorsal hand region. Our system achieves a mean joint-angle error of 8.81 degree for user-specific models and 9.77 degree for a general model. Further evaluation shows that our system outperforms previous work with an average of 20% higher accuracy in recognizing dynamic gestures, and achieves a 75% accuracy of detecting 11 different grasp types. We also demonstrate 3 applications which employ our system as a control device, an input device, and a grasped object recognizer.

We consider the task of re-calibrating the 3D pose of a static surveillance camera, whose pose may change due to external forces, such as birds, wind, falling objects or earthquakes. Conventionally, camera pose estimation can be solved with a PnP (Perspective-n-Point) method using 2D-to-3D feature correspondences, when 3D points are known. However, 3D point annotations are not always available or practical to obtain in real-world applications. We propose an alternative strategy for extracting 3D information to solve for camera pose by using pedestrian trajectories. We observe that 2D pedestrian trajectories indirectly contain useful 3D information that can be used for inferring camera pose. To leverage this information, we propose a data-driven approach by training a neural network (NN) regressor to model a direct mapping from 2D pedestrian trajectories projected on the image plane to 3D camera pose. We demonstrate that our regressor trained only on synthetic data can be directly applied to real data, thus eliminating the need to label any real data. We evaluate our method across six different scenes from the Town Centre Street and DUKEMTMC datasets. Our method achieves an improvement of ~ 50% on both position and orientation prediction accuracy when compared to other SOTA methods.

We address the problem of incrementally modeling and forecasting long-term goals of a first-person camera wearer: what the user will do, where they will go, and what goal they seek. In contrast to prior work in trajectory forecasting, our algorithm, Darko, goes further to reason about semantic states (will I pick up an object?), and future goal states that are far in terms of both space and time. Darko learns and forecasts from first-person visual observations of the user's daily behaviors via an Online Inverse Reinforcement Learning (IRL) approach. Classical IRL discovers only the rewards in a batch setting, whereas Darko discovers the transitions, rewards, and goals of a user from streaming data. Among other results, we show Darko forecasts goals better than competing methods in both noisy and ideal settings, and our approach is theoretically and empirically no-regret.

We propose two online-learning algorithms for modeling the personal preferences of users of interactive systems. The proposed algorithms leverage user feedback to estimate user behavior and provide personalized adaptive recommendation for supporting context-dependent decision-making. We formulate preference modeling as online prediction algorithms over a set of learned policies, i.e., policies generated via supervised learning with interaction and context data collected from previous users. The algorithms then adapt to a target user by learning the policy that best predicts that user’s behavior and preferences. We also generalize the proposed algorithms for a more challenging learning case in which they are restricted to a limited number of trained policies at each timestep, i.e., for mobile settings with limited resources. While the proposed algorithms are kept general for use in a variety of domains, we developed an image-filter-selection application. We used this application to demonstrate how the proposed algorithms can quickly learn to match the current user’s selections. Based on these evaluations, we show that (1) the proposed algorithms exhibit better prediction accuracy compared to traditional supervised learning and bandit algorithms, (2) our algorithms are robust under challenging limited prediction settings in which a smaller number of expert policies is assumed. Finally, we conducted a user study to demonstrate how presenting users with the prediction results of our algorithms significantly improves the efficiency of the overall interaction experience.

Navigation robots have the potential to overcome some of the limitations of traditional navigation aids for blind people, specially in unfamiliar environments. In this paper, we present the design of CaBot (Carry-on roBot), an autonomous suitcase-shaped navigation robot that is able to guide blind users to a destination while avoiding obstacles on their path. We conducted a user study where ten blind users evaluated specific functionalities of CaBot, such as a vibro-tactile handle to convey directional feedback; experimented to find their comfortable walking speed; and performed navigation tasks to provide feedback about their overall experience. We found that CaBot's performance highly exceeded users' expectations, who often compared it to navigating with a guide dog or sighted guide. Users' high confidence, sense of safety, and trust on CaBot poses autonomous navigation robots as a promising solution to increase the mobility and independence of blind people, in particular in unfamiliar environments.

People with visual impairments often have to rely on the assistance of sighted guides in airports, which prevents them from having an independent travel experience. In order to learn about their perspectives on current airport accessibility, we conducted two focus groups that discussed their needs and experiences in-depth, as well as the potential role of assistive technologies. We found that independent navigation is a main challenge and severely impacts their overall experience. As a result, we equipped an airport with a Bluetooth Low Energy (BLE) beacon-based navigation system and performed a real-world study where users navigated routes relevant for their travel experience. We found that despite the challenging environment participants were able to complete their itinerary independently, presenting none to few navigation errors and reasonable timings. This study presents the first systematic evaluation posing BLE technology as a strong approach to increase the independence of visually impaired people in airports.

We present an assistive suitcase system, BBeep, for supporting blind people when walking through crowded environments. BBeep uses pre-emptive sound notifications to help clear a path by alerting both the user and nearby pedestrians about the potential risk of collision. BBeep triggers notifications by tracking pedestrians, predicting their future position in real-time, and provides sound notifications only when it anticipates a future collision. We investigate how different types and timings of sound affect nearby pedestrian behavior. In our experiments, we found that sound emission timing has a significant impact on nearby pedestrian trajectories when compared to different sound types. Based on these findings, we performed a real-world user study at an international airport, where blind participants navigated with the suitcase in crowded areas. We observed that the proposed system significantly reduces the number of imminent collisions.

We propose the use of a proportional-derivative (PD) control based policy learned via reinforcement learning (RL) to estimate and forecast 3D human pose from egocentric videos. The method learns directly from unsegmented egocentric videos and motion capture data consisting of various complex human motions (e.g., crouching, hopping, bending, and motion transitions). We propose a video-conditioned recurrent control technique to forecast physically-valid and stable future motions of arbitrary length. We also introduce a value function based fail-safe mechanism which enables our method to run as a single pass algorithm over the video data. Experiments with both controlled and in-the-wild data show that our approach outperforms previous art in both quantitative metrics and visual quality of the motions, and is also robust enough to transfer directly to real-world scenarios. Additionally, our time analysis shows that the combined use of our pose estimation and forecasting can run at 30 FPS, making it suitable for real-time applications.

For autonomous vehicles (AVs) to behave appropriately on roads populated by human-driven vehicles, they must be able to reason about the uncertain intentions and decisions of other drivers from rich perceptual information. Towards these capabilities, we present a probabilistic forecasting model of future interactions between a variable number of agents. We perform both standard forecasting and the novel task of conditional forecasting, which reasons about how all agents will likely respond to the goal of a controlled agent (here, the AV). We train models on real and simulated data to forecast vehicle trajectories given past positions and LIDAR. Our evaluation shows that our model is substantially more accurate in multi-agent driving scenarios compared to existing state-of-the-art. Beyond its general ability to perform conditional forecasting queries, we show that our model's predictions of all agents improve when conditioned on knowledge of the AV's goal, further illustrating its capability to model agent interactions.

We study self-supervised adaptation of a robot's policy for social interaction, i.e., a policy for active communication with surrounding pedestrians through audio or visual signals. Inspired by the observation that humans continually adapt their behavior when interacting under varying social context, we propose Adaptive EXP4 (A-EXP4), a novel online learning algorithm for adapting the robot-pedestrian interaction policy. To address limitations of bandit algorithms in adaptation to unseen and highly dynamic scenarios, we employ a mixture model over the policy parameter space. Specifically, a Dirichlet Process Gaussian Mixture Model (DPMM) is used to cluster the parameters of sampled policies and maintain a mixture model over the clusters, hence effectively discovering policies that are suitable to the current environmental context in an unsupervised manner. Our simulated and real-world experiments demonstrate the feasibility of A-EXP4 in accommodating interaction with different types of pedestrians while jointly minimizing social disruption through the adaptation process. While the A-EXP4 formulation is kept general for application in a variety of domains requiring continual adaptation of a robot's policy, we specifically evaluate the performance of our algorithm using a suitcase-inspired assistive robotic platform. In this concrete assistive scenario, the algorithm observes how audio signals produced by the navigational system affect the behavior of pedestrians and adapts accordingly. Consequently, we find A-EXP4 to effectively adapt the interaction policy for gently clearing a navigation path in crowded settings, resulting in significant reduction in empirical regret compared to the EXP4 baseline.

We explore the possibility of using a single monocular camera to forecast the time to collision between a suitcase-shaped robot being pushed by its user and other nearby pedestrians. We develop a purely image-based deep learning approach that directly estimates the time to collision without the need of relying on explicit geometric depth estimates or velocity information to predict future collisions. While previous work has focused on detecting immediate collision in the context of navigating Unmanned Aerial Vehicles, the detection was limited to a binary variable (i.e., collision or no collision). We propose a more fine-grained approach to collision forecasting by predicting the exact time to collision in terms of milliseconds, which is more helpful for collision avoidance in the context of dynamic path planning. To evaluate our method, we have collected a novel dataset of over 13,000 indoor video segments each showing a trajectory of at least one person ending in a close proximity (a near collision) with the camera mounted on a mobile suitcase-shaped platform. Using this dataset, we do extensive experimentation on different temporal windows as input using an exhaustive list of state-of-the-art convolutional neural networks (CNNs). Our results show that our proposed multi-stream CNN is the best model for predicting time to near-collision. The average prediction error of our time to near-collision is 0.75 seconds across the test videos. The project webpage can be found at https://aashi7.github.io/NearCollision.html.

We focus on estimating the 3D orientation of the ground plane from a single image. We formulate the problem as an inter-mingled multi-task prediction problem by jointly optimizing for pixel-wise surface normal direction, ground plane segmentation, and depth estimates. Specifically, our proposed model, GroundNet, first estimates the depth and surface normal in two separate streams, from which two ground plane normals are then computed deterministically. To leverage the geometric correlation between depth and normal, we propose to add a consistency loss on top of the computed ground plane normals. In addition, a ground segmentation stream is used to isolate the ground regions so that we can selectively back-propagate parameter updates through only the ground regions in the image. Our method achieves the top-ranked performance on ground plane normal estimation and horizon line detection on the real-world outdoor datasets of ApolloScape and KITTI, improving the performance of previous art by up to 17.7% relatively.

The use of random perturbations of ground truth data, such as random translation or scaling of bounding boxes, is a common heuristic used for data augmentation that has been shown to prevent overfitting and improve generalization. Since the design of data augmentation is largely guided by reported best practices, it is difficult to understand if those design choices are optimal. To provide a more principled perspective, we develop a game-theoretic interpretation of data augmentation in the context of object detection. We aim to find an optimal adversarial perturbations of the ground truth data (i.e., the worst case perturbations) that forces the object bounding box predictor to learn from the hardest distribution of perturbed examples for better test-time performance. We establish that the game-theoretic solution (Nash equilibrium) provides both an optimal predictor and optimal data augmentation distribution. We show that our adversarial method of training a predictor can significantly improve test-time performance for the task of object detection. On the ImageNet, Pascal VOC and MS-COCO object detection tasks, our adversarial approach improves performance by about 16%, 5%, and 2% respectively compared to the best performing data augmentation methods.

We address the issue of domain gap when making use of synthetic data to train a scene-specific object detector and pose estimator. While previous works have shown that the constraints of learning a scene-specific model can be leveraged to create geometrically and photometrically consistent synthetic data, care must be taken to design synthetic content which is as close as possible to the real-world data distribution. In this work, we propose to solve domain gap through the use of appearance randomization to generate a wide range of synthetic objects to span the space of realistic images for training. An ablation study of our results is presented to delineate the individual contribution of different components in the randomization process. We evaluate our method on VIRAT, UA-DETRAC, EPFL-Car datasets, where we demonstrate that using scene specific domain randomized synthetic data is better than fine-tuning off-the-shelf models on limited real data.

To make images on Twitter and other social media platforms accessible to screen reader users, image descriptions (alternative text) need to be added that describe the information contained within the image. The lack of alternative text has been an enduring accessibility problem since the “alt” attribute was added in HTML 2.0 over 20 years ago, and the rise of user-generated content has only increased the number of images shared. As of 2016, Twitter provides users the ability to turn on a feature that allows descriptions to be added to images in their tweets, presumably in an effort to combat this accessibility problem. What has remained unknown is whether simply enabling users to provide alternative text has an impact on experienced accessibility. In this paper, we present a study of 1.09 million tweets with images, finding that only 0.1% of those tweets included descriptions. In a separate analysis of the timelines of 94 blind Twitter users, we found that these image tweets included descriptions more often. Even users with the feature turned on only write descriptions for about half of the images they tweet. To better understand why users provide alternative text descriptions (or not), we interviewed 20 Twitter users who have written image descriptions. Users did not remember to add alternative text, did not have time to add it, or did not know what to include when writing the descriptions. Our findings indicate that simply making it possible to provide image descriptions is not enough, and reveal future directions for automated tools that may support users in writing high-quality descriptions.

MOTIVATION: Since 2017, an increasing amount of attention has been paid to the supervised deep learning-based macromolecule in situ structural classification (i.e. subtomogram classification) in cellular electron cryo-tomography (CECT) due to the substantially higher scalability of deep learning. However, the success of such supervised approach relies heavily on the availability of large amounts of labeled training data. For CECT, creating valid training data from the same data source as prediction data is usually laborious and computationally intensive. It would be beneficial to have training data from a separate data source where the annotation is readily available or can be performed in a high-throughput fashion. However, the cross data source prediction is often biased due to the different image intensity distributions (a.k.a. domain shift). RESULTS: We adapt a deep learning-based adversarial domain adaptation (3D-ADA) method to timely address the domain shift problem in CECT data analysis. 3D-ADA first uses a source domain feature extractor to extract discriminative features from the training data as the input to a classifier. Then it adversarially trains a target domain feature extractor to reduce the distribution differences of the extracted features between training and prediction data. As a result, the same classifier can be directly applied to the prediction data. We tested 3D-ADA on both experimental and realistically simulated subtomogram datasets under different imaging conditions. 3D-ADA stably improved the cross data source prediction, as well as outperformed two popular domain adaptation methods. Furthermore, we demonstrate that 3D-ADA can improve cross data source recovery of novel macromolecular structures. AVAILABILITY AND IMPLEMENTATION: https://github.com/xulabs/projects. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

In many reinforcement learning problems, parameters of the model may vary with its phase while the agent attempts to learn through its interaction with the environment. For example, an autonomous car's reward on selecting a path may depend on traffic conditions at the time of the day or the transition dynamics of a drone may depend on the current wind direction. Many such processes exhibit a cyclic phase-structure and could be represented with a control policy parameterized over a circular or cyclic phase space. Attempting to model such phase variations with a standard data-driven approach (e.g. deep networks) without explicitly modeling the phase of the model can be challenging. Ambiguities may arise as the optimal action for a given state can vary depending on the phase. To better model cyclic environments, we propose phase-parameterized policies and value function approximators that explicitly enforce a cyclic structure to the policy or value space. We apply our phase-parameterized reinforcement learning approach to both feed-forward and recurrent deep networks in the context of trajectory optimization and locomotion problems. Our experiments show that our proposed approach has superior modeling performance than traditional function approximators in cyclic environments.

Feature representations from pre-trained deep neural networks have been known to exhibit excellent generalization and utility across a variety of related tasks. Fine-tuning is by far the simplest and most widely used approach that seeks to exploit and adapt these feature representations to novel tasks with limited data. Despite the effectiveness of fine-tuning, it is often sub-optimal and requires very careful optimization to prevent severe over-fitting to small datasets. The problem of sub-optimality and overfitting, is due in part to the large number of parameters used in a typical deep convolutional neural network. To address these problems, we propose a simple yet effective regularization method for fine-tuning pre-trained deep networks for the task of k-shot learning. To prevent overfitting, our key strategy is to cluster the model parameters while ensuring intra-cluster similarity and inter-cluster diversity of the parameters, effectively regularizing the dimensionality of the parameter search space. In particular, we identify groups of neurons within each layer of a deep network that shares similar activation patterns. When the network is to be fine-tuned for a classification task using only k examples, we propagate a single gradient to all of the neuron parameters that belong to the same group. The grouping of neurons is non-trivial as neuron activations depend on the distribution of the input data. To efficiently search for optimal groupings conditioned on the input data, we propose a reinforcement learning search strategy using recurrent networks to learn the optimal group assignments for each network layer. Experimental results show that our method can be easily applied to several popular convolutional neural networks and improve upon other state-of-the-art fine-tuning based k-shot learning strategies by more than 10%.

People with visual impairments (PVI) have shown interest in visiting museums and enjoying visual art. Based on this knowledge, some museums provide tactile reproductions of artworks, specialized tours for PVI, or enable them to schedule accessible visits. However, the ability of PVI to visit museums is still dependent on the assistance they get from their family and friends or from the museum personnel. In this paper, we surveyed 19 PVI to understand their opinions and expectations about visiting museums independently, as well as the requirements of user interfaces to support it. Moreover, we increase the knowledge about the previous experiences, motivations and accessibility issues of PVI in museums.

Indoor localization technologies can enhance quality of life for blind people by enabling them to independently explore and navigate indoor environments. Researchers typically evaluate their systems in terms of localization accuracy and user behavior along planned routes. We propose two measures of path-following behavior: deviation from optimal route and trajectory variability. Through regression analysis of real-world trajectories from blind users, we identify relationships between a) these measures and b) elements of the environment, route characteristics, localization error, and instructional cues that users receive. Our results provide insights into path-following behavior for turn-by-turn indoor navigation and have implications for the design of future interactions. Moreover, our findings highlight the importance of reporting these environmental factors and route properties in similar studies. We present automated and scalable methods for their calculation and to encourage their reporting for better interpretation and comparison of results across future studies.

Evaluating the impact of expertise and route knowledge on task performance can guide the design of intelligent and adaptive navigation interfaces. Expertise has been relatively unexplored in the context of assistive indoor navigation interfaces for blind people. To quantify the complex relationship between the user»s walking patterns, route learning, and adaptation to the interface, we conducted a study with 8 blind participants. The participants repeated a set of navigation tasks while using a smartphone-based turn-by-turn navigation guidance app. The results demonstrate the gradual evolution of user skill and knowledge throughout the route repetitions, significantly impacting the task completion time. In addition to the exploratory analysis, we take a step towards tailoring the navigation interface to the user»s needs by proposing a personalized recurrent neural network-based behavior model for expertise level classification.

Continuous and accurate smartphone-based localization is a promising technology for supporting independent mobility of people with visual impairments. However, despite extensive research on indoor localization techniques, they are still not ready for deployment in large and complex environments, like shopping malls and hospitals, where navigation assistance is needed. To achieve accurate, continuous, and real-time localization with smartphones in such environments, we present a series of key techniques enhancing a probabilistic localization algorithm. The algorithm is designed for smartphones and employs inertial sensors on a mobile device and Received Signal Strength (RSS) from Bluetooth Low Energy (BLE) beacons. We evaluate the proposed system in a 21,000 m2shopping mall which includes three multi-story buildings and a large open underground passageway. Experiments in this space validate the effect of the proposed technologies to improve localization accuracy. Field experiments with visually impaired participants confirm the practical performance of the proposed system in realistic use cases.

For many automated navigation applications, the underlying localization algorithm must be able to continuously produce both accurate and stable results by using a spectrum of redundant sensing technologies. To this end, various sensors have been used for localization, such as Wi-Fi, Bluetooth, GPS, LiDAR and cameras. In particular, a class of vision-based localization techniques using Structure from Motion (SfM) has been shown to produce very accurate position estimates in the real-world with moderate assumptions about the motion of the camera and the amount of visual texture in the environment. However, when these assumptions are violated, SfM techniques can fail catastrophically (i.e., cannot generate any estimate). Recently, a deep convolutional neural network (CNN) has been applied to images to robustly regress 6-DOF camera poses at the cost of lower accuracy than SfM. In this work, we propose improving image-based localization accuracy of deep CNN by combining Bluetooth radio-wave signal readings. In our experiments, we show that our proposed dual-stream CNN can robustly regress 6-DOF poses from images and radiowave signals better than one sensing modality alone. More importantly, we show that when both modes are used, the localization accuracy of the proposed deep CNN is comparable to that of SfM and significantly more robust than SfM.

Head movements are an integral part of human nonverbal communication. As such, the ability to detect various types of head gestures from video is important for robotic systems that need to interact with people or for assistive technologies that may need to detect conversational gestures to aid communication. To this end, we propose a novel Multi-Scale Deep Convolution-LSTM architecture, capable of recognizing short and long term motion patterns found in head gestures, from video data of natural and unconstrained conversations. In particular, our models use Convolutional Neural Networks (CNNs) to learn meaningful representations from short time windows over head motion data. To capture longer term dependencies, we use Recurrent Neural Networks (RNNs) that extract temporal patterns across the output of the CNNs. We compare against classical approaches using discriminative and generative graphical models and show that our model is able to significantly outperform baseline models.

Across most pedestrian detection datasets, it is typically assumed that pedestrians will be standing upright with respect to the image coordinate system. This assumption is not always valid for many vision-equipped mobile platforms, such as mobile phones, UAVs, or construction vehicles on rugged terrain. In these situations, the motion of the camera can cause images of pedestrians to be captured at extreme angles. This can lead to inferior pedestrian detection performance when using standard pedestrian detectors. To address this issue, we propose a Rotational Rectification Network (R2N) that can be inserted into any CNNbased pedestrian (or object) detector to adapt it to significant changes in camera rotation. The rotational rectification network uses a 2D rotation estimation module that passes rotational information to a spatial transformer network [12] to undistort image features. To enable robust rotation estimation, we propose a Global Polar Pooling (GPPooling) operator to capture rotational shifts in convolutional features. Through our experiments, we show how our rotational rectification network can be used to improve the performance of state-of-the-art pedestrian detectors under heavy image rotation by up to 45%.

We are interested in the development of image-based person detection algorithms for wearable computing using commodity smartphones. We focus on the use of smartphones as a wearable device because it is a practical means of augmenting human sensing for applications such as navigation for the blind or assisting social interaction. We identify two unique features of developing a vision-based person detector for body-worn smartphones: (1) the detector must take into account the strong bias in the size of people in the images taken with a wearable device and (2) the detector must consider the low image quality due to dim lighting and rapid ego-motion which leads to motion blur. In order to account for the unique distribution over the visibility of body parts when using a wearable camera, we propose a part-based person detector specialized for chestmounted smartphones. We perform extensive ablative analysis on the usefulness of part information, providing several insights regarding the design of the optimal person detector across different application domains. To account for the frequent occurrence of motion blur in our target domain, we introduce a data augmentation technique to generate synthetic motion-blurred images during training. In addition to addressing the aforementioned features, the final detector must also run in real-time using only smartphone resources. We leverage recent progress in deep neural networks for mobile devices and show that our proposed person detector, SmartPartNet, obtains performance similar to state-of-the-art pedestrian detection networks, while being 3X smaller and 5X faster.

We envision a future time when wearable cameras are worn by the masses and recording first-person point-of-view videos of everyday life. While these cameras can enable new assistive technologies and novel research challenges, they also raise serious privacy concerns. For example, first-person videos passively recorded by wearable cameras will necessarily include anyone who comes into the view of a camera-with or without consent. Motivated by these benefits and risks, we developed a self-search technique tailored to first-person videos. The key observation of our work is that the egocentric head motion of a target person (i.e., the self) is observed both in the point-of-view video of the target and observer. The motion correlation between the target person's video and the observer's video can then be used to identify instances of the self uniquely. We incorporate this feature into the proposed approach that computes the motion correlation over densely-sampled trajectories to search for a target individual in observer videos. Our approach significantly improves self-search performance over several well-known face detectors and recognizers. Furthermore, we show how our approach can enable several practical applications such as privacy filtering, target video retrieval, and social group clustering.

When preparing to visit new locations, sighted people often look at maps to build an a priori mental representation of the environment as a sequence of step-by-step actions and points of interest (POIs), e.g., turn right after the coffee shop. Based on this observation, we would like to understand if building the same type of sequential representation, prior to navigating in a new location, is helpful for people with visual impairments (VI). In particular, our goal is to understand how the simultaneous interplay between turn-by-turn navigation instructions and the relevant POIs in the route can aid the creation of a memorable sequential representation of the world. To this end, we present two smartphone-based virtual navigation interfaces: VirtualLeap, which allows the user to jump through a sequence of street intersection labels, turn-by-turn instructions and POIs along the route; and VirtualWalk, which simulates variable speed step-by-step walking using audio effects, whilst conveying similar route information. In a user study with 14 VI participants, most were able to create and maintain an accurate mental representation of both the sequential structure of the route and the approximate locations of the POIs. While both virtual navigation modalities resulted in similar spatial understanding, results suggests that each method is useful in different interaction contexts.

Delivering long term, unsupervised stroke rehabilitation in the home is a complex challenge that requires robust, low cost, scalable, and engaging solutions. We present HOMER, an interactive system that uses novel therapy artifacts, a computer vision approach, and a tablet interface to provide users with a flexible solution suitable for home based rehabilitation. HOMER builds on our prior work developing systems for lightly supervised rehabilitation use in the clinic, by identifying key features for functional movement analysis, adopting a simplified classification assessment approach, and supporting transferability of therapy outcomes to daily living experiences through the design of novel rehabilitation artifacts. A small pilot study with unimpaired subjects indicates the potential of the system in effectively assessing movement and establishing a creative environment for training.

Navigating in unfamiliar environments is challenging for most people, especially for individuals with visual impairments. While many personal navigation tools have been proposed to enable in- dependent indoor navigation, they have insufficient accuracy (e.g., 5-10 m), do not provide semantic features about surroundings (e.g., doorways, shops, etc.), and may require specialized devices to function. Moreover, the deployment of many systems is often only evaluated in constrained scenarios, which may not precisely reflect the performance in the real world. Therefore, we have de- signed and implemented NavCog3, a smartphone-based indoor navigation assistant that has been evaluated in a 21,000 m2 shop- ping mall. In addition to turn-by-turn instructions, it provides in- formation on landmarks (e.g., tactile paving) and points of interests nearby. We first conducted a controlled study with 10 visually im- paired users to assess localization accuracy and the perceived use- fulness of semantic features. To understand the usability of the app in a real-world setting, we then conducted another study with 43 participants with visual impairments where they could freely nav- igate in the shopping mall using NavCog3. Our findings suggest that NavCog3 can open a new opportunity for users with visual im- pairments to independently find and visit large and complex places with confidence.

Blind people often need to identify objects around them, from packages of food to items of clothing. Automatic object recognition continues to provide limited assistance in such tasks because models tend to be trained on images taken by sighted people with different background clutter, scale, viewpoints, occlusion, and image quality than in photos taken by blind users. We explore personal object recognizers, where visually impaired people train a mobile application with a few snapshots of objects of interest and provide custom labels. We adopt transfer learning with a deep learning system for user-defined multi-label k-instance classification. Experiments with blind participants demonstrate the feasibility of our approach, which reaches accuracies over 90% for some participants. We analyze user data and feedback to explore effects of sample size, photo-quality variance, and object shape; and contrast models trained on photos by blind participants to those by sighted participants and generic recognizers.

We develop predictive models of pedestrian dynamics by encoding the coupled nature of multi-pedestrian interaction using game theory and deep learning-based visual analysis to estimate person-specific behavior parameters. We focus on predictive models since they are important for developing interactive autonomous systems (e.g., autonomous cars, home robots, smart homes) that can understand different human behavior and pre-emptively respond to future human actions. Building predictive models for multi-pedestrian interactions however, is very challenging due to two reasons: (1) the dynamics of interaction are complex interdependent processes, where the decision of one person can affect others, and (2) dynamics are variable, where each person may behave differently (e.g., an older person may walk slowly while the younger person may walk faster). To address these challenges, we utilize concepts from game theory to model the intertwined decision making process of multiple pedestrians and use visual classifiers to learn a mapping from pedestrian appearance to behavior parameters. We evaluate our proposed model on several public multiple pedestrian interaction video datasets. Results show that our strategic planning model predicts and explains human interactions 25% better when compared to a state-of-the-art activity forecasting method.

State space models, such as Kalman filters or Particle filters, have been applied to improve the accuracy of radio-wave-based localization. However, these models can drift radically when assumptions of the models are violated, and they do not have a mechanism to fix errors. Therefore, we propose an approach to apply supervised learning to pedestrian localization, which is based on the Inference Machines framework. During training, we collect localization ground truths using computer vision while also collecting Bluetooth signals to train a state space model for localization, which can recover from model drift. During testing, our proposed approach uses only Bluetooth signals. Our experimental results show that our approach can improve the accuracy of Bluetooth-based localization with a small number of training examples. Moreover, our multi-modal supervision can also be used to estimate additional parameters, such as device rotation, from Bluetooth signals that do not have such information.

We address the problem of incrementally modeling and forecasting long-term goals of a first-person camera wearer: what the user will do, where they will go, and what goal they seek. In contrast to prior work in trajectory forecasting, our algorithm, DARKO, goes further to reason about semantic states (will I pick up an object?), and future goal states that are far in terms of both space and time. DARKO learns and forecasts from first-person visual observations of the user’s daily behaviors via an Online Inverse Reinforcement Learning (IRL) approach. Classical IRL discovers only the rewards in a batch setting, whereas DARKO discovers the states, transitions, rewards, and goals of a user from streaming data. Among other results, we show DARKO forecasts goals better than competing methods in both noisy and ideal settings, and our approach is theoretically and empirically no-regret.

We propose a privacy-preserving framework for learning visual classifiers by leveraging distributed private image data. This framework is designed to aggregate multiple classifiers updated locally using private data and to ensure that no private information about the data is exposed during and after its learning procedure. We utilize a homomorphic cryptosystem that can aggregate the local classifiers while they are encrypted and thus kept secret. To overcome the high computational cost of homomorphic encryption of high-dimensional classifiers, we (1) impose sparsity constraints on local classifier updates and (2) propose a novel efficient encryption scheme named doublypermuted homomorphic encryption (DPHE) which is tailored to sparse high-dimensional data. DPHE (i) decomposes sparse data into its constituent non-zero values and their corresponding support indices, (ii) applies homomorphic encryption only to the non-zero values, and (iii) employs double permutations on the support indices to make them secret. Our experimental evaluation on several public datasets shows that the proposed approach achieves comparable performance against state-of-the-art visual recognition methods while preserving privacy and significantly outperforms other privacy-preserving methods.

Given a retinal image, can we automatically determine whether it is of high quality (suitable for medical diagnosis)? Can we also explain our decision, pinpointing the region or regions that led to our decision? Images from human retinas are vital for the diagnosis of multiple health issues, like hypertension, diabetes, and Alzheimer's; low quality images may force the patient to come back again for a second scanning, wasting time and possibly delaying treatment. However, existing retinal image quality assessment methods are either black boxes without explanations of the results or depend heavily on feature engineering or on complex and error-prone anatomical structures' segmentation. Therefore, we propose EyeQual, that solves exactly this problem. EyeQual is novel, fast for inference, accurate and explainable, pinpointing low-quality regions on the image. We evaluated EyeQual on two real datasets where it achieved 100% accuracy taking just 36 milliseconds for each image.

Recurrent neural networks (RNNs) are a vital modeling technique that rely on internal states learned indirectly by optimization of a supervised, unsupervised, or reinforcement training loss. RNNs are used to model dynamic processes that are characterized by underlying latent states whose form is often unknown, precluding its analytic representation inside an RNN. In the Predictive-State Representation (PSR) literature, latent state processes are modeled by an internal state representation that directly models the distribution of future observations, and most recent work in this area has relied on explicitly representing and targeting sufficient statistics of this probability distribution. We seek to combine the advantages of RNNs and PSRs by augmenting existing state-of-the-art recurrent neural networks with Predictive-State Decoders (PSDs), which add supervision to the network's internal state representation to target predicting future observations. PSDs are simple to implement and easily incorporated into existing training pipelines via additional loss regularization. We demonstrate the effectiveness of PSDs with experimental results in three different domains: probabilistic filtering, Imitation Learning, and Reinforcement Learning. In each, our method improves statistical performance of state-of-the-art recurrent baselines and does so with fewer iterations and less data.

Great progress has been made in computer vision-based localization systems. However, some systems tend to work well only in certain visually feature-rich environments. It is often the case that feature-based matching techniques can have a hard time dealing with scenes with only a few features or a large number of repetitive features. In these situations, computer vision-based localization may fail to estimate camera position or may yield a large localization error. We approach this problem from a systems perspective, where we are required to obtain accurate localization of blind travellers using a smartphones app for localization. In particular, we assume that the environment is already instrumented with Bluetooth low energy (BLE) signals to provide rough proximity information, and we propose to integrate it with visual information to perform efficient structure-from-motion and camera localization. Our multi-model sensing approach can accelerate localization speed and obtain more accuracy in challenging environments when compared to traditional baseline approaches. We also show that our approach can accelerate the time for reconstructing large 3D models. Our framework is released as an open source project. It can be used by different mobile operating systems, enabling the development of navigation applications on mobile platforms.

This paper presents an egocentric vision (ego-vision) system for hand grasp analysis in unstructured environments. Our goal is to automatically recognize hand grasp types and to discover the visual structures of hand grasps using a wearable camera. In the proposed system, free hand–object interactions are recorded from a first-person viewing perspective. State-of-the-art computer vision techniques are used to detect hands and extract hand-based features. A new feature representation that incorporates hand tracking information is also proposed. Then, grasp classifiers are trained to discriminate among different grasp types from a predefined grasp taxonomy. Based on the trained grasp classifiers, visual structures of hand grasps are learned using an iterative grasp clustering method. In experiments, grasp recognition performance in both laboratory and real-world scenarios is evaluated. The best classification accuracy our system achieves is $\text{92}\%$ and $\text{59}\%$ , respectively. System generality to different tasks and users is also verified by the experiments. Analysis in a real-world scenario shows that it is possible to automatically learn intuitive visual grasp structures that are consistent with expert-designed grasp taxonomies.

The papers in this special section focus on the deployment of wearable computing technologies and ego-vision systems for augmented reality applications. Rapid progress in the development of low-level component technologies such as wearable sensors, wearable displays, and wearable computers is making our digital lives grow, connect, and play a relevant role in reality. To name a few examples, body-mounted sensors and displays help athletes in training by presenting real-time performance metrics such as speed, distance, and heart rate.Wearable systems allow medical staff in hospitals to consult specialists located anywhere in the world, in real time, providing optimal patient care. And within the context of assistive technologies, a head-mounted camera can be used to identify and convey the presence of objects, people, or text to a visually impaired user.

When navigating indoors, blind people are often unaware of key visual information, such as posters, signs, and exit doors. Our VizMap system uses computer vision and crowdsourcing to collect this information and make it available non-visually. VizMap starts with videos taken by on-site sighted volunteers and uses these to create a 3D spatial model. These video frames are semantically labeled by remote crowd workers with key visual information. These semantic labels are located within and embedded into the reconstructed 3D model, forming a query-able spatial representation of the environment. VizMap can then localize the user with a photo from their smartphone, and enable them to explore the visual elements that are nearby. We explore a range of example applications enabled by our reconstructed spatial representation. With VizMap, we move towards integrating the strengths of the end user, on-site crowd, online crowd, and computer vision to solve a long-standing challenge in indoor blind exploration.

We bring together ideas from recent work on feature design for egocentric action recognition under one framework by exploring the use of deep convolutional neural networks (CNN). Recent work has shown that features such as hand appearance, object attributes, local hand motion and camera ego-motion are important for characterizing first-person actions. To integrate these ideas under one framework, we propose a twin stream network architecture, where one stream analyzes appearance information and the other stream analyzes motion information. Our appearance stream encodes prior knowledge of the egocentric paradigm by explicitly training the network to segment hands and localize objects. By visualizing certain neuron activation of our network, we show that our proposed architecture naturally learns features that capture object attributes and hand-object configurations. Our extensive experiments on benchmark egocentric action datasets show that our deep architecture enables recognition rates that significantly outperform state-of-the-art techniques - an average 6:6% increase in accuracy over all datasets. Furthermore, by learning to recognize objects, actions and activities jointly, the performance of individual recognition tasks also increase by 30% (actions) and 14% (objects). We also include the results of extensive ablative analysis to highlight the importance of network design decisions.

When people observe and interact with physical spaces, they are able to associate functionality to regions in the environment. Our goal is to automate dense functional understanding of large spaces by leveraging sparse activity demonstrations recorded from an ego-centric viewpoint. The method we describe enables functionality estimation in large scenes where people have behaved, as well as novel scenes where no behaviors are observed. Our method learns and predicts "Action Maps", which encode the ability for a user to perform activities at various locations. With the usage of an egocentric camera to observe human activities, our method scales with the size of the scene without the need for mounting multiple static surveillance cameras and is well-suited to the task of observing activities up-close. We demonstrate that by capturing appearance-based attributes of the environment and associating these attributes with activity demonstrations, our proposed mathematical framework allows for the prediction of Action Maps in new environments. Additionally, we offer a preliminary glance of the applicability of Action Maps by demonstrating a proof-of concept application in which they are used in concert with activity detections to perform localization.

We aim to understand the dynamics of social interactions between two people by recognizing their actions and reactions using a head-mounted camera. Our work will impact several first-person vision tasks that need the detailed understanding of social interactions, such as automatic video summarization of group events and assistive systems. To recognize micro-level actions and reactions, such as slight shifts in attention, subtle nodding, or small hand actions, where only subtle body motion is apparent, we propose to use paired egocentric videos recorded by two interacting people. We show that the first-person and second-person points-of-view features of two people, enabled by paired egocentric videos, are complementary and essential for reliably recognizing micro-actions and reactions. We also build a new dataset of dyadic (two-persons) interactions that comprises more than 1000 pairs of egocentric videos to enable systematic evaluations on the task of micro-action and reaction recognition.

Turn-by-turn navigation is a useful paradigm for assisting people with visual impairments during mobility as it reduces the cognitive load of having to simultaneously sense, localize and plan. To realize such a system, it is necessary to be able to automatically localize the user with sufficient accuracy, provide timely and efficient instructions and have the ability to easily deploy the system to new spaces.

Where do we focus our attention in an image? Humans have an amazing ability to cut through the clutter to the parts of an image most relevant to the task at hand. Consider the task of geo-localizing tourist photos by retrieving other images taken at that location. Such photos naturally contain friends and family, and perhaps might even be nearly filled by a person's face if it is a selfie. Humans have no trouble ignoring these `distractions' and recognizing the parts that are indicative of location (e.g., the towers of Neuschwanstein Castle instead of their friend's face, a tree, or a car). In this paper, we investigate learning this ability automatically. At training-time, we learn how informative a region is for localization. At test-time, we use this learned model to determine what parts of a query image to use for retrieval. We introduce a new dataset, People at Landmarks, that contains large amounts of clutter in query images. Our system is able to outperform the existing state of the art approach to retrieval by more than 10% mAP, as well as improve results on a standard dataset without heavy occluders (Oxford5K).

Predicting the trajectory of a wide receiver in the game of American football requires prior knowledge about the game (e.g., route trees, defensive formations) and an accurate model of how the environment will change over time (e.g., opponent reaction strategies, motion attributes of players). Our aim is to build a computational model of the wide receiver, which takes into account prior knowledge about the game and short-term predictive models of how the environment will change over time. While prior knowledge of the game is readily accessible, it is quite challenging to build predictive models of how the environment will change over time. We propose several models for predicting short-term motions of opponent players to generate dynamic input features for our wide receiver forecasting model. In particular, we model the wide receiver with a Markov Decision Process (MDP), where the reward function is a linear combination of static features (prior knowledge about the game) and dynamic features (short-term prediction of opponent players). Since the dynamic features change over time, we make recursive calls to an inference procedure over the MDP while updating the dynamic features. We validate our technique on a video dataset of American football plays. Our results show that more informed models that accurately predict the motions of the defensive players are better at forecasting wide receiver plays.

We address the task of estimating large-scale land surface conditions using overhead aerial (macro-level) images and street view (micro-level) images. These two types of images are captured from orthogonal viewpoints and have different resolutions, thus conveying very different types of information that can be used in a complementary way. Moreover, their integration is necessary to enable an accurate understanding of changes in natural phenomena over massive city-scale landscapes. The key technical challenge is devising a method to integrate these two disparate types of image data in an effective manner, to leverage the wide coverage capabilities of macro-level images and detailed resolution of micro-level images. The strategy proposed in this work uses macro-level imaging to learn the extent to which the land condition corresponds between land regions that share similar visual characteristics (e.g., mountains, streets, buildings, rivers), whereas micro-level images are used to acquire high resolution statistics of land conditions (e.g., the amount of debris on the ground). By combining macro- and micro-level information about regional correspondences and surface conditions, our proposed method is capable of generating detailed estimates of land surface conditions over an entire city.

Maximum entropy inverse optimal control (MaxEnt IOC) is an effective means of discovering the underlying cost function of demonstrated human activity and can be used to predict human behavior over low-dimensional state spaces (i.e., forecasting of 2D trajectories). To enable inference in very large state spaces, we introduce an approximate MaxEnt IOC procedure to address the fundamental computational bottleneck stemming from calculating the partition function via dynamic programming. Approximate MaxEnt IOC is based on two components: approximate dynamic programming and Monte Carlo sampling. We analyze this approximation approach and provide a finite-sample error upper bound on its excess loss. We validate the proposed method in the context of analyzing dual-agent interactions from video, where we use approximate MaxEnt IOC to simulate mental images of a single agents body pose sequence (a high-dimensional image space). We experiment with sequences image data taken from RGB and RGBD data and show that it is possible to learn cost functions that lead to accurate predictions in high-dimensional problems that were previously intractable.

We consider the problem of designing a scene-specific pedestrian detector in a scenario where we have zero instances of real pedestrian data (i.e., no labeled real data or unsupervised real data). This scenario may arise when a new surveillance system is installed in a novel location and a scene-specific pedestrian detector must be trained prior to any observations of pedestrians. The key idea of our approach is to infer the potential appearance of pedestrians using geometric scene data and a customizable database of virtual simulations of pedestrian motion. We propose an efficient discriminative learning method that generates a spatially-varying pedestrian appearance model that takes into the account the perspective geometry of the scene. As a result, our method is able to learn a unique pedestrian classifier customized for every possible location in the scene. Our experimental results show that our proposed approach outperforms classical pedestrian detection models and hybrid synthetic-real models. Our results also yield a surprising result, that our method using purely synthetic data is able to outperform models trained on real scene-specific data when data is limited.

Our aim is to show how state-of-the-art computer vision techniques can be used to advance prehensile analysis (i.e., understanding the functionality of human hands). Prehensile analysis is a broad field of multi-disciplinary interest, where researchers painstakingly manually analyze hours of hand-object interaction videos to understand the mechanics of hand manipulation. In this work, we present promising empirical results indicating that wearable cameras and unsupervised clustering techniques can be used to automatically discover common modes of human hand use. In particular, we use a first-person point-of-view camera to record common manipulation tasks and leverage its strengths for reliably observing human hand use. To learn a diverse set of hand-object interactions, we propose a fast online clustering algorithm based on the Determinantal Point Process (DPP). Furthermore, we develop a hierarchical extension to the DPP clustering algorithm and show that it can be used to discover appearance-based grasp taxonomies. Using a purely data-driven approach, our proposed algorithm is able to obtain hand grasp taxonomies that roughly correspond to the classic Cutkosky grasp taxonomy. We validate our approach on over 10 hours of first-person point-of-view videos in both choreographed and real-life scenarios.

We envision a future time when wearable cameras (e.g., small cameras in glasses or pinned on a shirt collar) are worn by the masses and record first-person point-of-view (POV) videos of everyday life. While these cameras can enable new assistive technologies and novel research challenges, they also raise serious privacy concerns. For example, first-person videos passively recorded by wearable cameras will necessarily include anyone who comes into the view of a camera - with or without consent. Motivated by these benefits and risks, we develop a self-search technique tailored to first-person POV videos. The key observation of our work is that the egocentric head motions of a target person (i.e., the self) are observed both in the POV video of the target and observer. The motion correlation between the target person's video and the observer's video can then be used to uniquely identify instances of the self. We incorporate this feature into our proposed approach that computes the motion correlation over supervoxel hierarchies to localize target instances in observer videos. Our proposed approach significantly improves self-search performance over several well-known face detectors and recognizers. Furthermore, we show how our approach can enable several practical applications such as privacy filtering, automated video collection and social group discovery.

Wearable computing technologies are advancing rapidly and enabling users to easily record daily activities for applications such as life-logging or health monitoring. Recognizing hand and object interactions in these videos will help broaden application domains, but recognizing such interactions automatically remains a difficult task. Activity recognition from the first-person point-of-view is difficult because the video includes constant motion, cluttered backgrounds, and sudden changes of scenery. Recognizing hand-related activities is particularly challenging due to the many temporal and spatial variations induced by hand interactions. We present a novel approach to recognize hand-object interactions by extracting both local motion features representing the subtle movements of the hands and global hand shape features to capture grasp types. We validate our approach on multiple egocentric action datasets and show that state-of-the-art performance can be achieved by considering both local motion and global appearance information.

Our goal is to automatically recognize hand grasps and to discover the visual structures (relationships) between hand grasps using wearable cameras. Wearable cameras provide a first-person perspective which enables continuous visual hand grasp analysis of everyday activities. In contrast to previous work focused on manual analysis of first-person videos of hand grasps, we propose a fully automatic vision-based approach for grasp analysis. A set of grasp classifiers are trained for discriminating between different grasp types based on large margin visual predictors. Building on the output of these grasp classifiers, visual structures among hand grasps are learned based on an iterative discriminative clustering procedure. We first evaluated our classifiers on a controlled indoor grasp dataset and then validated the analytic power of our approach on real-world data taken from a machinist. The average F1 score of our grasp classifiers achieves over 0.80 for the indoor grasp dataset. Analysis of real-world video shows that it is possible to automatically learn intuitive visual grasp structures that are consistent with expert-designed grasp taxonomies.

We propose a novel method for performing fine-grained recognition of human hand grasp types using a single monocular image to allow computational systems to better understand human hand use. In particular, we focus on recognizing challenging grasp categories which differ only by subtle variations in finger configurations. While much of the prior work on understanding human hand grasps has been based on manual detection of grasps in video, this is the first work to automate the analysis process for fine-grained grasp classification. Instead of attempting to utilize a parametric model of the hand, we propose a hand parsing framework which leverages a data-driven learning to generate a pixel-wise segmentation of a hand into finger and palm regions. The proposed approach makes use of appearance-based cues such as finger texture and hand shape to accurately determine hand parts. We then build on the hand parsing result to compute high-level grasp features to learn a supervised fine-grained grasp classifier. To validate our approach, we introduce a grasp dataset recorded with a wearable camera, where the hand and its parts have been manually segmented with pixel-wise accuracy. Our results show that our proposed automatic hand parsing technique can improve grasp classification accuracy by over 30 percentage points over a state-of-the-art grasp recognition technique.

Achieving sub-pixel accuracy with face alignment algorithms is a difficult task given the diversity of appearance in real world facial profiles. To capture variations in perspective, occlusion, and illumination with adequate precision, current face alignment approaches rely on detecting facial landmarks and iteratively adjusting deformable models that encode prior knowledge of facial structure. However, these methods involve optimization in latent sub-spaces, where user-specific face shape information is easily lost after dimensionality reduction. Attempting to retain this information to capture this wide range of variation requires a large training distribution, which is difficult to obtain without high computational complexity. Subsequently, many face alignment methods lack the pixel-level accuracy necessary to satisfy the aesthetic requirements of tasks such as face deidentification, face swapping, and face modeling. In many such applications, the primary source of aesthetic inadequacy is a misaligned jaw line or facial contour. In this work, we explore the idea of an image-based refinement method to fix the landmark points of a misaligned facial contour. We propose an efficient two stage process - an intuitively constructed edge detection based algorithm to actively adjust facial contour landmark points, and a data driven validation system to filter out erroneous adjustments. Experimental results show that state-of-the-art face alignment combined with our proposed post-processing method yields improved overall performance over multiple face image datasets.

We address the task of pixel-level hand detection in the context of ego-centric cameras. Extracting hand regions in ego-centric videos is a critical step for understanding hand-object manipulation and analyzing hand-eye coordination. However, in contrast to traditional applications of hand detection, such as gesture interfaces or sign-language recognition, ego-centric videos present new challenges such as rapid changes in illuminations, significant camera motion and complex hand-object manipulations. To quantify the challenges and performance in this new domain, we present a fully labeled indoor/outdoor ego-centric hand detection benchmark dataset containing over 200 million labeled pixels, which contains hand images taken under various illumination conditions. Using both our dataset and a publicly available ego-centric indoors dataset, we give extensive analysis of detection performance using a wide range of local appearance features. Our analysis highlights the effectiveness of sparse features and the importance of modeling global illumination. We propose a modeling strategy based on our findings and show that our model outperforms several baseline approaches.

Egocentric cameras can be used to benefit such tasks as analyzing fine motor skills, recognizing gestures and learning about hand-object manipulation. To enable such technology, we believe that the hands must detected on the pixel-level to gain important information about the shape of the hands and fingers. We show that the problem of pixel-wise hand detection can be effectively solved, by posing the problem as a model recommendation task. As such, the goal of a recommendation system is to recommend the n-best hand detectors based on the probe set - a small amount of labeled data from the test distribution. This requirement of a probe set is a serious limitation in many applications, such as ego-centric hand detection, where the test distribution may be continually changing. To address this limitation, we propose the use of virtual probes which can be automatically extracted from the test distribution. The key idea is that many features, such as the color distribution or relative performance between two detectors, can be used as a proxy to the probe set. In our experiments we show that the recommendation paradigm is well-equipped to handle complex changes in the appearance of the hands in first-person vision. In particular, we show how our system is able to generalize to new scenarios by testing our model across multiple users.

We evaluate the performance of a widely used tracking-by-detection and data association multi-target tracking pipeline applied to an activity-rich video dataset. In contrast to traditional work on multi-target pedestrian tracking where people are largely assumed to be upright, we use an activity-rich dataset that includes a wide range of body poses derived from actions such as picking up an object, riding a bike, digging with a shovel, and sitting down. For each step of the tracking pipeline, we identify key limitations and offer practical modifications that enable robust multi-target tracking over a range of activities. We show that the use of multiple posture-specific detectors and an appearance-based data association post-processing step can generate non-fragmented trajectories essential for holistic activity understanding.

Portable high-quality sports cameras (e.g. head or helmet mounted) built for recording dynamic first-person video footage are becoming a common item among many sports enthusiasts. We address the novel task of discovering first-person action categories (which we call ego-actions) which can be useful for such tasks as video indexing and retrieval. In order to learn ego-action categories, we investigate the use of motion-based histograms and unsupervised learning algorithms to quickly cluster video content. Our approach assumes a completely unsupervised scenario, where labeled training videos are not available, videos are not pre-segmented and the number of ego-action categories are unknown. In our proposed framework we show that a stacked Dirichlet process mixture model can be used to automatically learn a motion histogram codebook and the set of ego-action categories. We quantitatively evaluate our approach on both in-house and public YouTube videos and demonstrate robust ego-action categorization across several sports genres. Comparative analysis shows that our approach outperforms other state-of-the-art topic models with respect to both classification accuracy and computational speed. Preliminary results indicate that on average, the categorical content of a 10 minute video sequence can be indexed in under 5 seconds.

In this paper we proposed a new interface for interacting with large displays via small video devices such as a cell phone. We estimate the location of the camera relative to the display using a matrix of transparent markers embedded on display. As a result, our interface allows the user to interact with digital contents without being distracted by opaque visual markers. Our interface enables intuitive interactions such as pointing, rotating, dragging and dropping. Moreover, our use of a small hand-held camera device allows for interaction with large scale displays without the need for direct contact with the display surface. Thus our system is well suited for interactions when there is some distance between the user and the display. Our proposed system has applications to large scale advertisement displays and can enable interactions between individuals and large scale digital content.

In this work, we propose a method for tracking individuals in crowds. Our method is based on a trajectory-based clustering approach that groups trajectories of image features that belong to the same person. The key novelty of our method is to make use of a person's individuality, that is, the gait features and the temporal consistency of local appearance to track each individual in a crowd. Gait features in the frequency domain have been shown to be an effective biometric cue in discriminating between individuals, and our method uses such features for tracking people in crowds for the first time. Unlike existing trajectory-based tracking methods, our method evaluates the dissimilarity of trajectories with respect to a group of three adjacent trajectories. In this way, we incorporate the temporal consistency of local patch appearance to differentiate trajectories of multiple people moving in close proximity. Our experiments show that the use of gait features and the temporal consistency of local appearance contributes to significant performance improvement in tracking people in crowded scenes.

The high-level recognition of human activity requires a priori hierarchical domain knowledge as well as a means of reasoning based on that knowledge. Based on insights from perceptual psychology, the problem of human action recognition is approached on the understanding that activities are hierarchical, temporally constrained and at times temporally overlapped. A hierarchical Bayesian network (HBN) based on a stochastic context-free grammar (SCFG) is implemented to address the hierarchical nature of human activity recognition. Then it is shown how the HBN is applied to different substrings in a sequence of primitive action symbols via deleted interpolation (DI) to recognize temporally overlapped activities. Results from the analysis of action sequences based on video surveillance data show the validity of the approach.

In recent years stochastic context-free grammars have been shown to be effective in modeling human activities because of the hierarchical structures they represent. However, most of the research in this area has yet to address the issue of learning the activity grammars from a noisy input source, namely, video. In this paper, we present a framework for identifying noise and recovering the basic activity grammar from a noisy symbol string produced by video. We identify the noise symbols by finding the set of non-noise symbols that optimally compresses the training data, where the optimality of compression is measured using an MDL criterion. We show the robustness of our system to noise and its effectiveness in learning the basic structure of human activity, through experiments with artificial data and a real video sequence from a local convenience store.