We address the critical gap between the computational demands of vision-language models and the possible ultra-low-bit weight precision (bitwidth

Most existing traffic video datasets including Waymo are structured, focusing predominantly on Western traffic, which hinders global applicability. Specifically, most Asian scenarios are far more complex, involving numerous objects with distinct motions and behaviors. Addressing this gap, we present a new dataset, DAVE, designed for evaluating perception methods with high representation of Vulnerable Road Users (VRUs: e.g. pedestrians, animals, motorbikes, and bicycles) in complex and unpredictable environments. DAVE is a manually annotated dataset encompassing 16 diverse actor categories (spanning animals, humans, vehicles, etc.) and 16 action types (complex and rare cases like cut-ins, zigzag movement, U-turn, etc.), which require high reasoning ability. DAVE densely annotates over 13 million bounding boxes (bboxes) actors with identification, and more than 1.6 million boxes are annotated with both actor identification and action/behavior details. The videos within DAVE are collected based on a broad spectrum of factors, such as weather conditions, the time of day, road scenarios, and traffic density. DAVE can benchmark video tasks like Tracking, Detection, Spatiotemporal Action Localization, Language-Visual Moment retrieval, and Multi-label Video Action Recognition. Given the critical importance of accurately identifying VRUs to prevent accidents and ensure road safety, in DAVE, vulnerable road users constitute 41.13% of instances, compared to 23.71% in Waymo. DAVE provides an invaluable resource for the development of more sensitive and accurate visual perception algorithms in the complex real world. Our experiments show that existing methods suffer degradation in performance when evaluated on DAVE, highlighting its benefit for future video recognition research.

Scene-aware Complementary Item Retrieval (CIR) is a challenging task which requires to generate a set of compatible items across domains. Due to the subjectivity, it is difficult to set up a rigorous standard for both data collection and learning objectives. To address this challenging task, we propose a visual compatibility concept, composed of similarity (resembling in color, geometry, texture, and etc.) and complementarity (different items like table vs chair completing a group). Based on this notion, we propose a compatibility learning framework, a category-aware Flexible Bidirectional Transformer (FBT), for visual ``scene-based set compatibility reasoning'' with the cross-domain visual similarity input and auto-regressive complementary item generation. We introduce a ``Flexible Bidirectional Transformer (FBT),'' consisting of an encoder with flexible masking, a category prediction arm, and an auto-regressive visual embedding prediction arm. And the inputs for FBT are cross-domain visual similarity invariant embeddings, making this framework quite generalizable. Furthermore, our proposed FBT model learns the inter-object compatibility from a large set of scene images in a self-supervised way. Compared with the SOTA methods, this approach achieves up to 5.3% and 9.6% in FITB score and 22.3% and 31.8% SFID improvement on fashion and furniture, respectively.

We introduce “HALLUSIONBENCH11“Hallusion” is a portmanteau of “hallucination” and “illusion.”,” a comprehensive benchmark designed for the evaluation of image-context rea-soning. This benchmark presents significant challenges to advanced large visual-language models (LVLMs), such as GPT-4V(ision), Gemini Pro Vision, Claude 3, and LLaVA-1.5, by emphasizing nuanced understanding and interpre-tation of visual data. The benchmark comprises 346 images paired with 1129 questions, all meticulously crafted by human experts. We introduce a novel structure for these visual questions designed to establish control groups. This structure enables us to conduct a quantitative analysis of the models' response tendencies, logical consistency, and various failure modes. In our evaluation on Hallusion-bench, we benchmarked 15 different models, highlighting a 31.42% question-pair accuracy achieved by the state-of-the-art GPT-4V. Notably, all other evaluated models achieve accuracy below 16%. Moreover, our analysis not only high-lights the observed failure modes, including language hal-lucination and visual illusion but also deepens an under-standing of these pitfalls. Our comprehensive case studies within Hallusionbench shed light on the challenges of hallucination and illusion in LVLMs. Based on these in-sights, we suggest potential pathways for their future im-provement. The benchmark and codebase can be accessed at https://github.com/tianyi-labIHallusionBench.

We present a new learning approach, Soft Conditional Prompt Learning (SCP), which leverages the strengths of prompt learning for aerial video action recognition. Our approach is designed to predict the action of each agent by helping the models focus on the descriptions or instructions associated with actions in the input videos for aerial/robot visual perception. Our formulation supports various prompts, including learnable prompts, auxiliary visual information, and large vision models to improve the recognition performance. We present a soft conditional prompt method that learns to dynamically generate prompts from a pool of prompt experts under different video inputs. By sharing the same objective with the task, our proposed SCP can optimize prompts that guide the model’s predictions while explicitly learning input-invariant (prompt experts pool) and input-specific (data-dependent) prompt knowledge. In practice, we observe a 3.17 − 10.2% accuracy improvement on the aerial video datasets (Okutama [1], NECDrone [2]), which consist of scenes with single-agent and multi-agent actions. We further evaluate our approach on ground camera videos to verify the effectiveness and generalization and achieve a 1.0 − 3.6% improvement on SSV2 [3]. We integrate our method into the ROS2 as well.

We present AGL-NET, a novel learning-based method for global localization using LiDAR point clouds and satellite maps. AGL-Net tackles two critical challenges: bridging the representation gap between image and points modalities for robust feature matching, and handling inherent scale discrepancies between global view and local view. To address these challenges, AGL-Net leverages a unified network architecture with a novel two-stage matching design. The first stage extracts informative neural features directly from raw sensor data and performs initial feature matching. The second stage refines this matching process by extracting informative skeleton features and incorporating a novel scale alignment step to rectify scale variations between LiDAR and map data. Furthermore, a novel scale and skeleton loss function guides the network toward learning scale-invariant feature representations, eliminating the need for pre-processing satellite maps. This significantly improves real-world applicability in scenarios with unknown map scales. To facilitate rigorous performance evaluation, we introduce a meticulously designed dataset within the CARLA simulator specifically tailored for metric localization training and assessment.

In trajectory forecasting tasks for traffic, future output trajectories can be computed by advancing the ego vehicle’s state with predicted actions according to a kinematics model. By unrolling predicted trajectories via time integration and models of kinematic dynamics, predicted trajectories should not only be kinematically feasible but also relate uncertainty from one timestep to the next. While current works in probabilistic prediction do incorporate kinematic priors for mean trajectory prediction, variance is often left as a learnable parameter, despite uncertainty in one time step being inextricably tied to uncertainty in the previous time step. In this paper, we show simple and differentiable analytical approximations describing the relationship between variance at one timestep and that at the next with the kinematic bicycle model. In our results, we find that encoding the relationship between variance across timesteps works especially well in unoptimal settings, such as with small or noisy datasets. We observe up to a 50% performance boost in partial dataset settings and up to an 8% performance boost in large-scale learning compared to previous kinematic prediction methods on SOTA trajectory forecasting architectures out-of-the-box, with no fine-tuning.

We present a new algorithm for the selection of informative frames in video action recognition. Our approach is designed for aerial videos captured using a moving camera where human actors occupy a small spatial resolution of video frames. Our algorithm utilizes the motion bias within aerial videos, which enables the selection of motion-salient frames. We introduce the concept of patch mutual information (PMI) score to quantify the motion bias between adjacent frames, by measuring the similarity of patches. We use this score to assess the amount of discriminative motion information contained in one frame relative to another. We present an adaptive frame selection strategy using shifted leaky ReLu and cumulative distribution function, which ensures that the sampled frames comprehensively cover all the essential segments with high motion salience. Our approach can be integrated with any action recognition model to enhance its accuracy. In practice, our method achieves a relative improvement of 2.2 - 13.8% in top-1 accuracy on UAV-Human, 6.8% on NEC Drone, and 9.0% on Diving48 datasets. The code is available at https://github.com/Ricky-Xian/PMI-Sampler.

We present a novel approach for action recognition in UAV videos. Our formulation is designed to handle occlusion and viewpoint changes caused by the movement of a UAV. We use the concept of mutual information to compute and align the regions corresponding to human action or motion in the temporal domain. This enables our recognition model to learn from the key features associated with the motion. We also propose a novel frame sampling method that uses joint mutual information to acquire the most informative frame sequence in UAV videos. We have integrated our approach with X3D and evaluated the performance on multiple datasets. In practice, we achieve 18.9% improvement in Top-1 accuracy over current state-of-the-art methods on UAV-Human [30], 7.3% improvement on Drone-Action [41], and 7.16% improvement on NEC Drones [7]. The code is available at https://github.com/Ricky-Xian/MITFAS

We present CrossLoc3D, a novel 3D place recognition method that solves a large-scale point matching problem in a cross-source setting. Cross-source point cloud data corresponds to point sets captured by depth sensors with different accuracies or from different distances and perspectives. We address the challenges in terms of developing 3D place recognition methods that account for the representation gap between points captured by different sources. Our method handles cross-source data by utilizing multi-grained features and selecting convolution kernel sizes that correspond to most prominent features. Inspired by the diffusion models, our method uses a novel iterative refinement process that gradually shifts the embedding spaces from different sources to a single canonical space for better metric learning. In addition, we present CS-Campus3D, the first 3D aerial-ground cross-source dataset consisting of point cloud data from both aerial and ground LiDAR scans. The point clouds in CS-Campus3D have representation gaps and other features like different views, point densities, and noise patterns. We show that our CrossLoc3D algorithm can achieve an improvement of 4.74% - 15.37% in terms of the top 1 average recall on our CS-Campus3D benchmark and achieves performance comparable to state-of-the-art 3D place recognition method on the Oxford RobotCar. The code and CS-Campus3D benchmark will be available at github.com/rayguan97/crossloc3d.

Driven by the need from real-world applications, Auxiliary Modality Learning (AML) offers the possibility to utilize more information from auxiliary data in training, while only requiring data from one or fewer modalities in test, to save the overall computational cost and reduce the amount of input data for inferencing. In this work, we formally define ``Auxiliary Modality Learning'' (AML), systematically classify types of auxiliary modality (in visual computing) and architectures for AML, and analyze their performance. We also analyze the conditions under which AML works well from the optimization and data distribution perspectives. To guide various choices to achieve optimal performance using AML, we propose a novel method to assist in choosing the best auxiliary modality and estimating an upper bound performance before executing AML. In addition, we propose a new AML method using generalized curriculum distillation to enable more effective curriculum learning. Our method achieves the best performance compared to other SOTA methods.

We present a new traffic dataset, Meteor, which captures traffic patterns and multi-agent driving behaviors in unstructured scenarios. Meteor consists of more than 1000 one-minute videos, over 2 million annotated frames with bounding boxes and GPS trajectories for 16 unique agent categories, and more than 13 million bounding boxes for traffic agents. Meteor is a dataset for rare and interesting, multi-agent driving behaviors that are grouped into traffic violations, atypical interactions, and diverse scenarios. Every video in Meteor is tagged using a diverse range of factors corresponding to weather, time of the day, road conditions, and traffic density. We use Meteor to benchmark perception methods for object detection and multi-agent behavior prediction. Our key finding is that state-of-the-art models for object detection and behavior prediction, which otherwise succeed on existing datasets such as Waymo, fail on the Meteor dataset. Meteor is a step towards developing more sophisticated perception models for dense, heterogeneous, and unstructured scenarios.

In this paper, we propose a novel learning framework for autonomous systems that uses a small amount of “auxiliary information” that complements the learning of the main modality, called “small-shot auxiliary modality distillation network (AMD-S-Net)”. The AMD-S-Net contains a two-stream framework design that can fully extract information from different types of data (i.e., paired/unpaired multi-modality data) to distill knowledge more effectively. We also propose a novel training paradigm based on the “reset operation” that enables the teacher to explore the local loss landscape near the student domain iteratively, providing local landscape information and potential directions to discover better solutions by the student, thus achieving higher learning performance. Our experiments show that AMD-S-Net and our training paradigm outperform other SOTA methods by up to 12.7% and 18.1% improvement in autonomous steering, respectively.

We propose a novel approach for aerial video action recognition. Our method is designed for videos captured using UAVs and can run on edge or mobile devices. We present a learning-based approach that uses customized auto zoom to automatically identify the human target and scale it appropriately. This makes it easier to extract the key features and reduces the computational overhead. We also present an efficient temporal reasoning algorithm to capture the action information along the spatial and temporal domains within a controllable computational cost. Our approach has been implemented and evaluated both on the desktop with high-end GPUs and on the low power Robotics RB5 Platform for robots and drones. In practice, we achieve$6.1-7.4 \%$improvement over SOTA in Top-1 accuracy on the RoCoG-v2 dataset, 8.3-10.4% improvement on the UAV-Human dataset and 3.2% improvement on the Drone Action dataset.

We propose a new convolution called Dynamic Region-Aware Convolution (DRConv), which can automatically assign multiple filters to corresponding spatial regions where features have similar representation. In this way, DR-Conv outperforms standard convolution in modeling semantic variations. Standard convolutional layer can increase the number of filers to extract more visual elements but results in high computational cost. More gracefully, our DR-Conv transfers the increasing channel-wise filters to spatial dimension with learnable instructor, which not only improve representation ability of convolution, but also maintains computational cost and the translation-invariance as standard convolution dose. DRConv is an effective and elegant method for handling complex and variable spatial information distribution. It can substitute standard convolution in any existing networks for its plug-and-play property, especially to power convolution layers in efficient networks. We evaluate DRConv on a wide range of models (MobileNet series, ShuffleNetV2, etc.) and tasks (Classification, Face Recognition, Detection and Segmentation). On ImageNet classification, DRConv-based ShuffleNetV2-0.5 achieves state-of-the-art performance of 67.1% at 46M× multiply-adds level with 6.3% relative improvement.

Benefitted from its great success on many tasks, deep learning is increasingly used on low-computational-cost devices, e.g. smartphone, embedded devices, etc. To reduce the high computational and memory cost, in this work, we propose a fully learnable group convolution module (FLGC for short) which is quite efficient and can be embedded into any deep neural networks for acceleration. Specifically, our proposed method automatically learns the group structure in the training stage in a fully end-to-end manner, leading to a better structure than the existing pre-defined, two-steps, or iterative strategies. Moreover, our method can be further combined with depthwise separable convolution, resulting in 5 times acceleration than the vanilla Resnet50 on single CPU. An additional advantage is that in our FLGC the number of groups can be set as any value, but not necessarily 2^k as in most existing methods, meaning better tradeoff between accuracy and speed. As evaluated in our experiments, our method achieves better performance than existing learnable group convolution and standard group convolution when using the same number of groups.