Classifier-free guidance has become a staple for conditional generation with denoising diffusion models. However, a comprehensive understanding of classifier-free guidance is still missing. In this work, we carry out an empirical study to provide a fresh perspective on classifier-free guidance. Concretely, instead of solely focusing on classifier-free guidance, we trace back to the root, i.e., classifier guidance, pinpoint the key assumption for the derivation, and conduct a systematic study to understand the role of the classifier. On 1D data, we find that both classifier guidance and classifier-free guidance achieve conditional generation by pushing the denoising diffusion trajectories away from decision boundaries, i.e., areas where conditional information is usually entangled and is hard to learn. To validate this classifier-centric perspective on high-dimensional data, we assess whether a flow-matching postprocessing step that is designed to narrow the gap between a pre-trained diffusion model’s learned distribution and the real data distribution, especially near decision boundaries, can improve the performance. Experiments on various datasets verify our classifier-centric understanding.

We propose to synthesize high-quality and synchronized audio, given video and optional text conditions, using a novel multimodal joint training framework (MMAudio). In contrast to single-modality training conditioned on (limited) video data only, MMAudio is jointly trained with larger-scale, readily available text-audio data to learn to generate semantically aligned high-quality audio samples. Additionally, we improve audio-visual synchrony with a conditional synchronization module that aligns video conditions with audio latents at the frame level. Trained with a flow matching objective, MMAudio achieves new video-to-audio state-of-the-art among public models in terms of audio quality, semantic alignment, and audio-visual synchronization, while having a low inference time (1.23s to generate an 8s clip) and just 157M parameters. MMAudio also achieves surprisingly competitive performance in text-to-audio generation, showing that joint training does not hinder single-modality performance. Code, models, and demo are available at: hkchengrex.github.io/MMAudio.

We present Relocate, a simple training-free baseline designed to perform the challenging task of visual query localization in long videos. To eliminate the need for task-specific training and efficiently handle long videos, Relocate leverages a region-based representation derived from pretrained vision models. At a high level, it follows the classic object localization approach: (1) identify all objects in each video frame, (2) compare the objects with the given query and select the most similar ones, and (3) perform bidirectional tracking to get a spatio-temporal response. However, we propose some key enhancements to handle small objects, cluttered scenes, partial visibility, and varying appearances. Notably, we refine the selected objects for accurate localization and generate additional visual queries to capture visual variations. We evaluate Relocate on the challenging Ego4D Visual Query 2D Localization dataset, establishing a new baseline that outperforms prior task-specific methods by 49% (relative improvement) in spatio-temporal average precision.

Recent sparse multi-view scene reconstruction advances like DUSt3R and MASt3R no longer require camera calibration and camera pose estimation. However, they only process a pair of views at a time to infer pixel-aligned pointmaps. When dealing with more than two views, a combinatorial number of error prone pairwise reconstructions are usually followed by an expensive global optimization, which often fails to rectify the pairwise reconstruction errors. To handle more views, reduce errors, and improve inference time, we propose the fast single-stage feed-forward network MV- DUSt3R. At its core are multi-view decoder blocks which exchange information across any number of views while considering one reference view. To make our method robust to reference view selection, we further propose MV-DUSt3R+, which employs cross-reference-view blocks to fuse information across different reference view choices. To further enable novel view synthesis, we extend both by adding and jointly training Gaussian splatting heads. Experiments on multi-view stereo reconstruction, multi-view pose estimation, and novel view synthesis confirm that our methods improve significantly upon prior art. Code released.1

Minibatch optimal transport coupling straightens paths in unconditional flow matching. This leads to computationally less demanding inference as fewer integration steps and less complex numerical solvers can be employed when numerically solving an ordinary differential equation at test time. However, in the conditional setting, minibatch optimal transport falls short. This is because the default optimal transport mapping disregards conditions, resulting in a conditionally skewed prior distribution during training. In contrast, at test time, we have no access to the skewed prior, and instead sample from the full, unbiased prior distribution. This gap between training and testing leads to a subpar performance. To bridge this gap, we propose conditional optimal transport C^2OT that adds a conditional weighting term in the cost matrix when computing the optimal transport assignment. Experiments demonstrate that this simple fix works with both discrete and continuous conditions in 8gaussians-to-moons, CIFAR-10, ImageNet-32x32, and ImageNet-256x256. Our method performs better overall compared to the existing baselines across different function evaluation budgets. Code is available at https://hkchengrex.github.io/C2OT

Generalizable rendering of an animatable human avatar from sparse inputs relies on data priors and inductive biases extracted from training on large data to avoid scene-specific optimization and to enable fast reconstruction. This raises two main challenges: First, unlike iterative gradient-based adjustment in scene-specific optimization, generalizable methods must reconstruct the human shape representation in a single pass at inference time. Second, rendering is preferably computationally efficient yet of high resolution. To address both challenges we augment the recently proposed dual shape representation, which combines the benefits of a mesh and Gaussian points, in two ways. To improve reconstruction, we propose an iterative feedback update framework, which successively improves the canonical human shape representation during reconstruction. To achieve computationally efficient yet high-resolution rendering, we study a coupled-multi-resolution Gaussians-on-Mesh representation. We evaluate the proposed approach on the challenging THuman2.0, XHuman and AIST++ data. Our approach reconstructs an animatable representation from sparse inputs in less than 1s, renders views with 95.1FPS at $1024 \times 1024$, and achieves PSNR/LPIPS*/FID of 24.65/110.82/51.27 on THuman2.0, outperforming the state-of-the-art in rendering quality.

We formulate a hierarchical rectified flow to model data distributions. It hierarchically couples multiple ordinary differential equations (ODEs) and defines a time-differentiable stochastic process that generates a data distribution from a known source distribution. Each ODE resembles the ODE that is solved in a classic rectified flow, but differs in its domain, i.e., location, velocity, acceleration, etc. Unlike the classic rectified flow formulation, which formulates a single ODE in the location domain and only captures the expected velocity field (sufficient to capture a multi-modal data distribution), the hierarchical rectified flow formulation models the multi-modal random velocity field, acceleration field, etc., in their entirety. This more faithful modeling of the random velocity field enables integration paths to intersect when the underlying ODE is solved during data generation. Intersecting paths in turn lead to integration trajectories that are more straight than those obtained in the classic rectified flow formulation, where integration paths cannot intersect. This leads to modeling of data distributions with fewer neural function evaluations. We empirically verify this on synthetic 1D and 2D data as well as MNIST, CIFAR-10, and ImageNet-32 data. Our code is available at: https://riccizz.github.io/HRF/.

We study Variational Rectified Flow Matching, a framework that enhances classic rectified flow matching by modeling multi-modal velocity vector-fields. At inference time, classic rectified flow matching ’moves’ samples from a source distribution to the target distribution by solving an ordinary differential equation via integration along a velocity vector-field. At training time, the velocity vector-field is learnt by linearly interpolating between coupled samples one drawn from the source and one drawn from the target distribution randomly. This leads to ”ground-truth” velocity vector-fields that point in different directions at the same location, i.e., the velocity vector-fields are multi-modal/ambiguous. However, since training uses a standard mean-squared-error loss, the learnt velocity vector-field averages ”ground-truth” directions and isn’t multi-modal. In contrast, variational rectified flow matching learns and samples from multi-modal flow directions. We show on synthetic data, MNIST, CIFAR-10, and ImageNet that variational rectified flow matching leads to compelling results.

Recent work studying the generalization of diffusion models with locally linear UNet-based denoisers reveals inductive biases that can be expressed via geometry-adaptive harmonic bases. For such locally linear UNets, these geometry-adaptive harmonic bases can be conveniently visualized through the eigen-decomposition of a UNet’s Jacobian matrix. In practice, however, more recent denoising networks are often transformer-based, e.g., the diffusion transformer (DiT). Due to the presence of nonlinear operations, similar eigen-decomposition analyses cannot be used to reveal the inductive biases of transformer-based denoisers. This motivates our search for alternative ways to explain the strong generalization ability observed in DiT models. Investigating a DiT’s pivotal attention modules, we find that locality of attention maps in a DiT’s early layers are closely associated with generalization. To verify this finding, we modify the generalization of a DiT by restricting its attention windows and observe an improvement in generalization. Furthermore, we empirically find that both the placement and the effective attention size of these local attention windows are crucial factors. Experimental results on the CelebA, ImageNet, MSCOCO, and LSUN data show that strengthening the inductive bias of a DiT can improve both generalization and generation quality when less training data is available. Source code is available at https://github.com/DiT-Generalization/DiT-Generalization.

We introduce the Region Encoder Network (REN), a fast and effective model for generating region-based image representations using point prompts. Recent methods combine class-agnostic segmenters (e.g., SAM) with patch-based image encoders (e.g., DINO) to produce compact and effective region representations, but they suffer from high computational cost due to the segmentation step. REN bypasses this bottleneck using a lightweight module that directly generates region tokens, enabling 60x faster token generation with 35x less memory, while also improving token quality. It uses a few cross-attention blocks that take point prompts as queries and features from a patch-based image encoder as keys and values to produce region tokens that correspond to the prompted objects. We train REN with three popular encoders—DINO, DINOv2, and OpenCLIP—and show that it can be extended to other encoders without dedicated training. We evaluate REN on semantic segmentation and retrieval tasks, where it consistently outperforms the original encoders in both performance and compactness, and matches or exceeds SAM-based region methods while being significantly faster. Notably, REN achieves state-of-the-art results on the challenging Ego4D VQ2D benchmark and outperforms proprietary LMMs on Visual Haystacks' single-needle challenge. The code and pretrained models are available at https://github.com/savya08/ren.

We tackle the task of recovering an animatable 3D human avatar from a single or a sparse set of images. For this task, beyond a set of images, many prior state-of-the-art methods use accurate “ground-truth” camera poses and human poses as input to guide reconstruction at test-time. We show that pose‑dependent reconstruction degrades results significantly if pose estimates are noisy. To overcome this, we introduce NoPo-Avatar, which reconstructs avatars solely from images, without any pose input. By removing the dependence of test-time reconstruction on human poses, NoPo-Avatar is not affected by noisy human pose estimates, making it more widely applicable. Experiments on challenging THuman2.0, XHuman, and HuGe100K data show that NoPo-Avatar outperforms existing baselines in practical settings (without ground‑truth poses) and delivers comparable results in lab settings (with ground‑truth poses).

Backpropagation, which uses the chain rule, is the de-facto standard algorithm for optimizing neural networks nowadays. Recently, Hinton (2022) proposed the forward-forward algorithm, a promising alternative that optimizes neural nets layer-by-layer, without propagating gradients throughout the network. Although such an approach has several advantages over back-propagation and shows promising results, the fact that each layer is being trained independently limits the optimization process. Specifically, it prevents the network's layers from collaborating to learn complex and rich features. In this work, we study layer collaboration in the forward-forward algorithm. We show that the current version of the forward-forward algorithm is suboptimal when considering information flow in the network, resulting in a lack of collaboration between layers of the network. We propose an improved version that supports layer collaboration to better utilize the network structure, while not requiring any additional assumptions or computations. We empirically demonstrate the efficacy of the proposed version when considering both information flow and objective metrics. Additionally, we provide a theoretical motivation for the proposed method, inspired by functional entropy theory.

We present Cutie, a video object segmentation (VOS) network with object-level memory reading, which puts the object representation from memory back into the video object segmentation result. Recent works on VOS employ bottom-up pixel-level memory reading which struggles due to matching noise, especially in the presence of distractors, resulting in lower performance in more challenging data. In contrast, Cutie performs top-down object-level memory reading by adapting a small set of object queries. Via those, it interacts with the bottom-up pixel features iteratively with a query-based object transformer (qt, hence Cutie). The object queries act as a high-level summary of the target object, while high-resolution feature maps are retained for accurate segmentation. Together with foreground-background masked attention, Cutie cleanly separates the semantics of the foreground object from the background. On the challenging MOSE dataset, Cutie improves by 8.7$\mathcal{J}\& \mathcal{F}$over XMem with a similar running time and improves by 4.2$\mathcal{J}\&\mathcal{F}$over DeAOT while being three times faster. Code is available at: hkchengrex.github.io/Cutie.

We present a method for automatically modifying a NeRF representation based on a single observation of a non-rigid transformed version of the original scene. Our method defines the transformation as a 3D flow, specifically as a weighted linear blending of rigid transformations of 3D anchor points that are defined on the surface of the scene. In order to identify anchor points, we introduce a novel correspondence algorithm that first matches RGB-based pairs, then leverages multi-view information and 3D reprojection to robustly filter false positives in two steps. We also introduce a new dataset for exploring the problem of modifying a NeRF scene through a single observation. Our dataset11https://nerfdeformer.github.io/ contains 113 synthetic scenes leveraging 47 3D assets. We show that our proposed method outperforms NeRF editing methods as well as diffusion-based methods, and we also explore different methods for filtering correspondences.

We introduce GoMAvatar, a novel approach for real-time, memory-efficient, high-quality animatable human modeling. GoMAvatar takes as input a single monocular video to create a digital avatar capable of re-articulation in new poses and real-time rendering from novel view-points, while seamlessly integrating with rasterization-based graphics pipelines. Central to our method is the Gaussians-on-Mesh (GoM) representation, a hybrid 3D model combining rendering quality and speed of Gaussian splatting with geometry modeling and compatibility of deformable meshes. We assess GoMAvatar on ZJU-MoCap, PeopleSnapshot, and various YouTube videos. GoMAvatar matches or surpasses current monocular human modeling algorithms in rendering quality and significantly outperforms them in computational efficiency (43 FPS) while being memory-efficient (3.63 MB per subject).

Rendering scenes observed in a monocular video from novel viewpoints is a challenging problem. For static scenes the community has studied both scene-specific optimization techniques, which optimize on every test scene, and generalized techniques, which only run a deep net forward pass on a test scene. In contrast, for dynamic scenes, scene-specific optimization techniques exist, but, to our best knowledge, there is currently no generalized method for dynamic novel view synthesis from a given monocular video. To explore whether generalized dynamic novel view synthesis from monocular videos is possible today, we establish an analysis framework based on existing techniques and work toward the generalized approach. We find a pseudo-generalized process without scene-specific \emph{appearance} optimization is possible, but geometrically and temporally consistent depth estimates are needed. Despite no scene-specific appearance optimization, the pseudo-generalized approach improves upon some scene-specific methods.For more information see project page at https://xiaoming-zhao.github.io/projects/pgdvs.

Protein design, a grand challenge of the day, involves optimization on a fitness landscape, and leading methods adopt a model-based approach where a model is trained on a training set (protein sequences and fitness) and proposes candidates to explore next. These methods are challenged by sparsity of high-fitness samples in the training set, a problem that has been in the literature. A less recognized but equally important problem stems from the distribution of training samples in the design space: leading methods are not designed for scenarios where the desired optimum is in a region that is not only poorly represented in training data, but also relatively far from the highly represented low-fitness regions. We show that this problem of “separation” in the design space is a significant bottleneck in existing model-based optimization tools and propose a new approach that uses a novel VAE as its search model to overcome the problem. We demonstrate its advantage over prior methods in robustly finding improved samples, regardless of the imbalance and separation between low- and high-fitness samples. Our comprehensive benchmark on real and semi-synthetic protein datasets as well as solution design for physics-informed neural networks, showcases the generality of our approach in discrete and continuous design spaces. Our implementation is available at https://github.com/sabagh1994/PGVAE.

In real-world scenarios, arbitrary interactions with the environment can often be costly, and actions of expert demonstrations are not always available. To reduce the need for both, offline Learning from Observations (LfO) is extensively studied: the agent learns to solve a task given only expert states and task-agnostic non-expert state-action pairs. The state-of-the-art DIstribution Correction Estimation (DICE) methods, as exemplified by SMODICE, minimize the state occupancy divergence between the learner’s and empirical expert policies. However, such methods are limited to either $f$-divergences (KL and $\chi^2$) or Wasserstein distance with Rubinstein duality, the latter of which constrains the underlying distance metric crucial to the performance of Wasserstein-based solutions. To enable more flexible distance metrics, we propose Primal Wasserstein DICE (PW-DICE). It minimizes the primal Wasserstein distance between the learner and expert state occupancies and leverages a contrastively learned distance metric. Theoretically, our framework is a generalization of SMODICE, and is the first work that unifies $f$-divergence and Wasserstein minimization. Empirically, we find that PW-DICE improves upon several state-of-the-art methods. The code is available at https://github.com/KaiYan289/PW-DICE.

We propose the new task open-world video instance segmentation and captioning. It requires to detect, segment, track and describe with rich captions never before seen objects. This challenging task can be addressed by developing "abstractors" which connect a vision model and a language foundation model. Concretely, we connect a multi-scale visual feature extractor and a large language model (LLM) by developing an object abstractor and an object-to-text abstractor. The object abstractor, consisting of a prompt encoder and transformer blocks, introduces spatially-diverse open-world object queries to discover never before seen objects in videos. An inter-query contrastive loss further encourages the diversity of object queries. The object-to-text abstractor is augmented with masked cross-attention and acts as a bridge between the object queries and a frozen LLM to generate rich and descriptive object-centric captions for each detected object. Our generalized approach surpasses the baseline that jointly addresses the tasks of open-world video instance segmentation and dense video object captioning by 13% on never before seen objects, and by 10% on object-centric captions.

Decision Transformers have recently emerged as a new and compelling paradigm for offline Reinforcement Learning (RL), completing a trajectory in an autoregressive way. While improvements have been made to overcome initial shortcomings, online finetuning of decision transformers has been surprisingly under-explored. The widely adopted state-of-the-art Online Decision Transformer (ODT) still struggles when pretrained with low-reward offline data. In this paper, we theoretically analyze the online-finetuning of the decision transformer, showing that the commonly used Return-To-Go (RTG) that's far from the expected return hampers the online fine-tuning process. This problem, however, is well-addressed by the value function and advantage of standard RL algorithms. As suggested by our analysis, in our experiments, we hence find that simply adding TD3 gradients to the finetuning process of ODT effectively improves the online finetuning performance of ODT, especially if ODT is pretrained with low-reward offline data. These findings provide new directions to further improve decision transformers.

Single-view RGB-D human reconstruction with implicit functions is often formulated as per-point classification. Specifically, a set of 3D locations within the view-frustum of the camera are first projected independently onto the image and a corresponding feature is subsequently extracted for each 3D location. The feature of each 3D location is then used to classify independently whether the corresponding 3D point is inside or outside the observed object. This procedure leads to sub-optimal results because correlations between predictions for neighboring locations are only taken into account implicitly via the extracted features. For more accurate results we propose the occupancy planes (OPlanes) representation, which enables to formulate single-view RGB-D human reconstruction as occupancy prediction on planes which slice through the camera's view frustum. Such a representation provides more flexibility than voxel grids and enables to better leverage correlations than per-point classification. On the challenging S3D data we observe a simple classifier based on the OPlanes representation to yield compelling results, especially in difficult situations with partial occlusions due to other objects and partial visibility, which haven't been addressed by prior work.

Real world images often have highly imbalanced content density. Some areas are very uniform, e.g., large patches of blue sky, while other areas are scattered with many small objects. Yet, the commonly used successive grid downsampling strategy in convolutional deep networks treats all areas equally. Hence, small objects are represented in very few spatial locations, leading to worse results in tasks such as segmentation. Intuitively, retaining more pixels representing small objects during downsampling helps to preserve important information. To achieve this, we propose AutoFocusFormer (AFF), a local-attention transformer image recognition backbone, which performs adaptive down-sampling by learning to retain the most important pixels for the task. Since adaptive downsampling generates a set of pixels irregularly distributed on the image plane, we abandon the classic grid structure. Instead, we develop a novel point-based local attention block, facilitated by a balanced clustering module and a learnable neighborhood merging module, which yields representations for our point-based versions of state-of-the-art segmentation heads. Experiments show that our AutoFocusFormer (AFF) improves significantly over baseline models of similar sizes.

In this work, we present a novel framework built to sim-plify 3D asset generation for amateur users. To enable interactive generation, our method supports a variety of input modalities that can be easily provided by a human, in-cluding images, text, partially observed shapes and combinations of these, further allowing to adjust the strength of each input. At the core of our approach is an encoder-decoder, compressing 3D shapes into a compact latent representation, upon which a diffusion model is learned. To enable a variety of multimodal inputs, we employ task-specific encoders with dropout followed by a cross-attention mechanism. Due to its flexibility, our model naturally supports a variety of tasks, outperforming prior works on shape completion, image-based 3D reconstruction, and text-to-3D. Most interestingly, our model can combine all these tasks into one swiss-army-knife tool, enabling the user to perform shape generation using incomplete shapes, images, and textual descriptions at the same time, providing the relative weights for each input and facilitating interactivity. Despite our approach being shape-only, we further show an efficient method to texture the generated shape using large-scale text-to-image models.

Object queries have emerged as a powerful abstraction to generically represent object proposals. However, their use for temporal tasks like video segmentation poses two questions: 1) How to process frames sequentially and propagate object queries seamlessly across frames. Using independent object queries per frame doesn't permit tracking, and requires post-processing. 2) How to produce temporally consistent, yet expressive object queries that model both appearance and position changes. Using the entire video at once doesn't capture position changes and doesn't scale to long videos. As one answer to both questions we propose 'context-aware relative object queries', which are continuously propagated frame-by-frame. They seamlessly track objects and deal with occlusion and re-appearance of objects, without post-processing. Further, we find context-aware relative object queries better capture position changes of objects in motion. We evaluate the proposed approach across three challenging tasks: video instance segmentation, multi-object tracking and segmentation, and video panoptic segmentation. Using the same approach and architecture, we match or surpass state-of-the art results on the diverse and challenging OVIS, Youtube-VIS, Cityscapes-VPS, MOTS 2020 and KITTI-MOTS data.

Training data for video segmentation are expensive to annotate. This impedes extensions of end-to-end algorithms to new video segmentation tasks, especially in large-vocabulary settings. To ‘track anything’ without training on video data for every individual task, we develop a decoupled video segmentation approach (DEVA), composed of task-specific image-level segmentation and class/task-agnostic bi-directional temporal propagation. Due to this design, we only need an image-level model for the target task (which is cheaper to train) and a universal temporal propagation model which is trained once and generalizes across tasks. To effectively combine these two modules, we use bi-directional propagation for (semi-)online fusion of segmentation hypotheses from different frames to generate a coherent segmentation. We show that this decoupled formulation compares favorably to end-to-end approaches in several data-scarce tasks including large-vocabulary video panoptic segmentation, open-world video segmentation, referring video segmentation, and unsupervised video object segmentation. Code is available at: hkchengrex.github.io/Tracking-Anything-with-DEVA.

Reconstructing the 3D shape of objects observed in a single image is a challenging task. Recent approaches rely on visual cues extracted from a given image learned from a deep net. In this work, we leverage recent advances in monocular scene understanding to incorporate an additional geometric cue of surface normals. For this, we proposed a novel optimization layer that encourages the face normals of the reconstructed shape to be aligned with estimated surface normals. We develop a computationally efficient conjugate-gradient-based method that avoids the computation of a high-dimensional sparse matrix. We show this framework to achieve compelling shape reconstruction results on the challenging Pix3D and ShapeNet datasets.

We present a system for collision-free control of a robot manipulator that uses only RGB views of the world. Perceptual input of a tabletop scene is provided by multiple images of an RGB camera (without depth) that is either handheld or mounted on the robot end effector. A NeRF-like process is used to reconstruct the 3D geometry of the scene, from which the Euclidean full signed distance function (ESDF) is computed. A model predictive control algorithm is then used to control the manipulator to reach a desired pose while avoiding obstacles in the ESDF. We show results on a real dataset collected and annotated in our lab. Our results are also available at https://ngp-mpc.github.io/.

Offline imitation from observations aims to solve MDPs where only task-specific expert states and task-agnostic non-expert state-action pairs are available. Offline imitation is useful in real-world scenarios where arbitrary interactions are costly and expert actions are unavailable. The state-of-the-art ‘DIstribution Correction Estimation’ (DICE) methods minimize divergence of state occupancy between expert and learner policies and retrieve a policy with weighted behavior cloning; however, their results are unstable when learning from incomplete trajectories, due to a non-robust optimization in the dual domain. To address the issue, in this paper, we propose Trajectory-Aware Imitation Learning from Observations (TAILO). TAILO uses a discounted sum along the future trajectory as the weight for weighted behavior cloning. The terms for the sum are scaled by the output of a discriminator, which aims to identify expert states. Despite simplicity, TAILO works well if there exist trajectories or segments of expert behavior in the task-agnostic data, a common assumption in prior work. In experiments across multiple testbeds, we find TAILO to be more robust and effective, particularly with incomplete trajectories.

in "A Realism Metric for Generated LiDAR Point Clouds" by Triess et al. In this work, a new metric is proposed that measures the quality of LiDAR point clouds that are generated, e.g., by a generative network. The metric can be used as an early indicator which assesses whether the generated training data will improve a down-stream task like semantic point cloud segmentation. Brissman et al. propose in "Recurrent Graph Neural Networks for Video Instance Segmentation" a very efficient approach for video instance segmentation. The approach tracks and segments multiple objects on-line and in real-time. Temporal information is also used in "Spatio-Temporal Outdoor Lighting Aggregation on Image Sequences using Transformer Networks" by Lee et al. to estimate outdoor illumination more consistently and without the need of an additional post-processing step.

Recent work on 3D-aware image synthesis has achieved compelling results using advances in neural rendering. However, 3D-aware synthesis of face dynamics hasn't received much attention. Here, we study how to explicitly control generative model synthesis of face dynamics exhibiting non-rigid motion (e.g., facial expression change), while simultaneously ensuring 3D-awareness. For this we propose a Controllable Radiance Field (CoRF): 1) Motion control is achieved by embedding motion features within the layered latent motion space of a style-based generator; 2) To ensure consistency of background, motion features and subject-specific attributes such as lighting, texture, shapes, albedo, and identity, a face parsing net, a head regressor and an identity encoder are incorporated. On head image/video data we show that CoRFs are 3D-aware while enabling editing of identity, viewing directions, and motion.

Recently, there has been an increasing interest in building question answering (QA) models that reason across multiple modalities, such as text and images. However, QA using images is often limited to just picking the answer from a pre-defined set of options. In addition, images in the real world, especially in news, have objects that are co-referential to the text, with complementary information from both modalities. In this paper, we present a new QA evaluation benchmark with 1,384 questions over news articles that require cross-media grounding of objects in images onto text. Specifically, the task involves multi-hop questions that require reasoning over image-caption pairs to identify the grounded visual object being referred to and then predicting a span from the news body text to answer the question. In addition, we introduce a novel multimedia data augmentation framework, based on cross-media knowledge extraction and synthetic question-answer generation, to automatically augment data that can provide weak supervision for this task. We evaluate both pipeline-based and end-to-end pretraining-based multimedia QA models on our benchmark, and show that they achieve promising performance, while considerably lagging behind human performance hence leaving large room for future work on this challenging new task.

Designing equivariance as an inductive bias into deep-nets has been a prominent approach to build effective models, e.g., a convolutional neural network incorporates translation equivariance. However, incorporating these inductive biases requires knowledge about the equivariance properties of the data, which may not be available, e.g., when encountering a new domain. To address this, we study how to "discover interpretable equivariances" from data. Specifically, we formulate this discovery process as an optimization problem over a model’s parameter-sharing schemes. We propose to use the partition distance to empirically quantify the accuracy of the recovered equivariance. Also, we theoretically analyze the method for Gaussian data and provide a bound on the mean squared gap between the studied discovery scheme and the oracle scheme. Empirically, we show that the approach recovers known equivariances, such as permutations and shifts, on sum of numbers and spatially-invariant data.

Image segmentation groups pixels with different semantics, e.g., category or instance membership. Each choice of semantics defines a task. While only the semantics of each task differ, current research focuses on designing spe-cialized architectures for each task. We present Masked- attention Mask Transformer (Mask2Former), a new archi-tecture capable of addressing any image segmentation task (panoptic, instance or semantic). Its key components in-clude masked attention, which extracts localized features by constraining cross-attention within predicted mask regions. In addition to reducing the research effort by at least three times, it outperforms the best specialized architectures by a significant margin on four popular datasets. Most no-tably, Mask2Former sets a new state-of-the-art for panoptic segmentation (57.8 PQ on COCO), instance segmentation (50.1 AP on COCO) and semantic segmentation (57.7 mIoU onADE20K).

Forecasting of a representation is important for safe and effective autonomy. For this, panoptic segmentations have been studied as a compelling representation in recent work. However, recent state-of-the-art on panoptic segmentation forecasting suffers from two issues: first, individual object instances are treated independently of each other; second, individual object instance forecasts are merged in a heuristic manner. To address both issues, we study a new panoptic segmentation forecasting model that jointly forecasts all object instances in a scene using a transformer model based on ‘difference attention.’ It further refines the predictions by taking depth estimates into account. We evaluate the proposed model on the Cityscapes and AIODrive datasets. We find difference attention to be particularly suitable for forecasting because the difference of quantities like locations enables a model to explicitly reason about velocities and acceleration. Because of this, we attain state-of-the-art on panoptic segmentation forecasting metrics.

We introduce an approach for selecting objects in neural volumetric 3D representations, such as multi-plane images (MPI) and neural radiance fields (NeRF). Our approach takes a set of foreground and background 2D user scribbles in one view and automatically estimates a 3D segmentation of the desired object, which can be rendered into novel views. To achieve this result, we propose a novel voxel feature embedding that incorporates the neural volumetric 3D representation and multi-view image features from all input views. To evaluate our approach, we introduce a new dataset of human-provided segmentation masks for depicted objects in real-world multi-view scene captures. We show that our approach out-performs strong baselines, including 2D segmentation and 3D segmentation approaches adapted to our task.

Optimization within a layer of a deep-net has emerged as a new direction for deep-net layer design. However, there are two main challenges when applying these layers to computer vision tasks: (a) which optimization problem within a layer is useful?; (b) how to ensure that computation within a layer remains efficient? To study question (a), in this work, we propose total variation (TV) minimization as a layer for computer vision. Motivated by the success of total variation in image processing, we hypothesize that TV as a layer provides useful inductive bias for deep-nets too. We study this hypothesis on five computer vision tasks: image classification, weakly supervised object localization, edge-preserving smoothing, edge detection, and image denoising, improving over existing baselines. To achieve these results we had to address question (b): we developed a GPU-based projected-Newton method which is 37× faster than existing solutions.

To solve difficult tasks, humans ask questions to acquire knowledge from external sources. In contrast, classical reinforcement learning agents lack such an ability and often resort to exploratory behavior. This is exacerbated as few present-day environments support querying for knowledge. In order to study how agents can be taught to query external knowledge via language, we first introduce two new environments: the grid-world-based Q-BabyAI and the text-based Q-TextWorld. In addition to physical interactions, an agent can query an external knowledge source specialized for these environments to gather information. Second, we propose the ‘Asking for Knowledge’ (AFK) agent, which learns to generate language commands to query for meaningful knowledge that helps solve the tasks. AFK leverages a non-parametric memory, a pointer mechanism and an episodic exploration bonus to tackle (1) irrelevant information, (2) a large query language space, (3) delayed reward for making meaningful queries. Extensive experiments demonstrate that the AFK agent outperforms recent baselines on the challenging Q-BabyAI and Q-TextWorld environments.

We address the problem of visual storytelling, i.e., generating a story for a given sequence of images. While each story sentence should describe a corresponding image, a coherent story also needs to be consistent and relate to both future and past images. Current approaches encode images independently, disregarding relations between images. Our approach learns to encode images with different interactions based on the story position (i.e., past image or future image). To this end, we develop a novel message-passing-like algorithm for ordered image attention (OIA) that collects interactions across all the images in the sequence. Finally, to generate the story's sentences, a second attention mechanism picks the important image attention vectors with an Image-Sentence Attention (ISA). The obtained results improve the METEOR score on the VIST dataset by 1%. Furthermore, a thorough human study confirms improvements and demonstrates that order-based interactions significantly improve coherency (64.20% \vs 28.70%). Source code available at \urlhttps://github.com/tomateb/OIAVist.git

Generative adversarial nets (GANs) have been remarkably successful at learning to sample from distributions specified by a given dataset, particularly if the given dataset is reasonably large compared to its dimensionality. However, given limited data, classical GANs have struggled, and strategies like output-regularization, data-augmentation, use of pre-trained models and pruning have been shown to lead to improvements. Notably, the applicability of these strategies is often constrained to particular settings, e.g., availability of a pretrained GAN, or increases training time, e.g., when using pruning. In contrast, we propose a Discriminator gradIent Gap regularized GAN (DigGAN) formulation which can be added to any existing GAN. DigGAN augments existing GANs by encouraging to narrow the gap between the norm of the gradient of a discriminator's prediction w.r.t. real images and w.r.t. the generated samples. We observe this formulation to avoid bad attractors within the GAN loss landscape, and we find DigGAN to significantly improve the results of GAN training when limited data is available.

Generalization bounds which assess the difference between the true risk and the empirical risk have been studied extensively. However, to obtain bounds, current techniques use strict assumptions such as a uniformly bounded or a Lipschitz loss function. To avoid these assumptions, in this paper, we follow an alternative approach: we relax uniform bounds assumptions by using on-average bounded loss and on-average bounded gradient norm assumptions. Following this relaxation, we propose a new generalization bound that exploits the contractivity of the log-Sobolev inequalities. These inequalities add an additional loss-gradient norm term to the generalization bound, which is intuitively a surrogate of the model complexity. We apply the proposed bound on Bayesian deep nets and empirically analyze the effect of this new loss-gradient norm term on different neural architectures.

We propose learnable polyphase sampling (LPS), a pair of learnable down/upsampling layers that enable truly shift-invariant and equivariant convolutional networks. LPS can be trained end-to-end from data and generalizes existing handcrafted downsampling layers. It is widely applicable as it can be integrated into any convolutional network by replacing down/upsampling layers. We evaluate LPS on image classification and semantic segmentation. Experiments show that LPS is on-par with or outperforms existing methods in both performance and shift consistency. For the first time, we achieve true shift-equivariance on semantic segmentation (PASCAL VOC), i.e., 100% shift consistency, outperforming baselines by an absolute 3.3%.

Although reinforcement learning has found widespread use in dense reward settings, training autonomous agents with sparse rewards remains challenging. To address this difficulty, prior work has shown promising results when using not only task-specific demonstrations but also task-agnostic albeit somewhat related demonstrations. In most cases, the available demonstrations are distilled into an implicit prior, commonly represented via a single deep net. Explicit priors in the form of a database that can be queried have also been shown to lead to encouraging results. To better benefit from available demonstrations, we develop a method to Combine Explicit and Implicit Priors (CEIP). CEIP exploits multiple implicit priors in the form of normalizing flows in parallel to form a single complex prior. Moreover, CEIP uses an effective explicit retrieval and push-forward mechanism to condition the implicit priors. In three challenging environments, we find the proposed CEIP method to improve upon sophisticated state-of-the-art techniques.

Our goal is to forecast the near future given a set of recent observations. We think this ability to forecast, i.e., to anticipate, is integral for the success of autonomous agents which need not only passively analyze an observation but also must react to it in real-time. Importantly, accurate forecasting hinges upon the chosen scene decomposition. We think that superior forecasting can be achieved by decomposing a dynamic scene into individual ‘things’ and background ‘stuff’. Background ‘stuff’ largely moves because of camera motion, while foreground ‘things’ move because of both camera and individual object motion. Following this decomposition, we introduce panoptic segmentation forecasting. Panoptic segmentation forecasting opens up a middle-ground between existing extremes, which either forecast instance trajectories or predict the appearance of future image frames. To address this task we develop a two-component model: one component learns the dynamics of the background stuff by anticipating odometry, the other one anticipates the dynamics of detected things. We establish a leaderboard for this novel task, and validate a state-of-the-art model that outperforms available baselines.

Extracting detailed 3D information of objects from video data is an important goal for holistic scene understanding. While recent methods have shown impressive results when reconstructing meshes of objects from a single image, results often remain ambiguous as part of the object is unobserved. Moreover, existing image-based datasets for mesh reconstruction don’t permit to study models which integrate temporal information. To alleviate both concerns we present SAIL-VOS 3D: a synthetic video dataset with frame-by-frame mesh annotations which extends SAIL-VOS. We also develop first baselines for reconstruction of 3D meshes from video data via temporal models. We demonstrate efficacy of the proposed baseline on SAIL-VOS 3D and Pix3D, showing that temporal information improves reconstruction quality. Resources and additional information are available at http://sailvos.web.illinois.edu.

We introduce WyPR, a Weakly-supervised framework for Point cloud Recognition, requiring only scene-level class tags as supervision. WyPR jointly addresses three core 3D recognition tasks: point-level semantic segmentation, 3D proposal generation, and 3D object detection, coupling their predictions through self and cross-task consistency losses. We show that in conjunction with standard multiple-instance learning objectives, WyPR can detect and segment objects in point cloud data without access to any spatial labels at training time. We demonstrate its efficacy using the ScanNet and S3DIS datasets, outperforming prior state of the art on weakly-supervised segmentation by more than 6% mIoU. In addition, we set up the first benchmark for weakly-supervised 3D object detection on both datasets, where WyPR outperforms standard approaches and establishes strong baselines for future work.

Multi-object tracking and segmentation (MOTS) is important for understanding dynamic scenes in video data. Existing methods perform well on multi-object detection and segmentation for independent video frames, but tracking of objects over time remains a challenge. MOTS methods formulate tracking locally, i.e., frame-by-frame, leading to sub-optimal results. Classical global methods on tracking operate directly on object detections, which leads to a combinatorial growth in the detection space. In contrast, we formulate a global method for MOTS over the space of assignments rather than detections: First, we find all top-k assignments of objects detected and segmented between any two consecutive frames and develop a structured prediction formulation to score assignment sequences across any number of consecutive frames. We use dynamic programming to find the global optimizer of this formulation in polynomial time. Second, we connect objects which reappear after having been out of view for some time. For this we formulate an assignment problem. On the challenging KITTI-MOTS and MOTSChallenge datasets, this achieves state-of-the-art results among methods which don’t use depth data.

While deep reinforcement learning (RL) promises freedom from hand-labeled data, great successes, especially for Embodied AI, require significant work to create supervision via carefully shaped rewards. Indeed, without shaped rewards, i.e., with only terminal rewards, present-day Embodied AI results degrade significantly across Embodied AI problems from single-agent Habitat-based PointGoal Navigation (SPL drops from 55 to 0) and two-agent AI2-THOR-based Furniture Moving (success drops from 58% to 1%) to three-agent Google Football-based 3 vs. 1 with Keeper (game score drops from 0.6 to 0.1). As training from shaped rewards doesn’t scale to more realistic tasks, the community needs to improve the success of training with terminal rewards. For this we propose GRIDTOPIX: 1) train agents with terminal rewards in gridworlds that generically mirror Embodied AI environments, i.e., they are independent of the task; 2) distill the learned policy into agents that reside in complex visual worlds. Despite learning from only terminal rewards with identical models and RL algorithms, GRIDTOPIX significantly improves results across tasks: from PointGoal Navigation (SPL improves from 0 to 64) and Furniture Moving (success improves from 1% to 25%) to football gameplay (game score improves from 0.1 to 0.6). GRIDTOPIX even helps to improve the results of shaped reward training.

Communication between embodied AI agents has received increasing attention in recent years. Despite its use, it is still unclear whether the learned communication is interpretable and grounded in perception. To study the grounding of emergent forms of communication, we first introduce the collaborative multi-object navigation task ‘CoMON.' In this task, an ‘oracle agent' has detailed environment information in the form of a map. It communicates with a ‘navigator agent' that perceives the environment visually and is tasked to find a sequence of goals. To succeed at the task, effective communication is essential. CoMON hence serves as a basis to study different communication mechanisms between heterogeneous agents, that is, agents with different capabilities and roles. We study two common communication mechanisms and analyze their communication patterns through an egocentric and spatial lens. We show that the emergent communication can be grounded to the agent observations and the spatial structure of the 3D environment.

It is fundamental for personal robots to reliably navigate to a specified goal. To study this task, PointGoal navigation has been introduced in simulated Embodied AI environments. Recent advances solve this PointGoal navigation task with near-perfect accuracy (99.6% success) in photo-realistically simulated environments, assuming noiseless egocentric vision, noiseless actuation and most importantly, perfect localization. However, under realistic noise models for visual sensors and actuation, and without access to a "GPS and Compass sensor," the 99.6%-success agents for PointGoal navigation only succeed with 0.3%.1In this work, we demonstrate the surprising effectiveness of visual odometry for the task of PointGoal navigation in this realistic setting, i.e., with realistic noise models for perception and actuation and without access to GPS and Compass sensors. We show that integrating visual odometry techniques into navigation policies improves the state-of-the-art on the popular Habitat PointNav benchmark by a large margin, improving success from 64.5% to 71.7% while executing 6.4 times faster.

Exploration is critical for good results in deep reinforcement learning and has attracted much attention. However, existing multi-agent deep reinforcement learning algorithms still use mostly noise-based techniques. Very recently, exploration methods that consider cooperation among multiple agents have been developed. However, existing methods suffer from a common challenge: agents struggle to identify states that are worth exploring, and hardly coordinate exploration efforts toward those states. To address this shortcoming, in this paper, we propose cooperative multi-agent exploration (CMAE): agents share a common goal while exploring. The goal is selected from multiple projected state spaces by a normalized entropy-based technique. Then, agents are trained to reach the goal in a coordinated manner. We demonstrate that CMAE consistently outperforms baselines on various tasks, including a sparse-reward version of multiple-particle environment (MPE) and the Starcraft multi-agent challenge (SMAC).

Solving complex real-world tasks, e.g., autonomous fleet control, often involves a coordinated team of multiple agents which learn strategies from visual inputs via reinforcement learning. Many existing multi-agent reinforcement learning (MARL) algorithms however don’t scale to environments where agents operate on visual inputs. To address this issue, algorithmically, recent works have focused on non-stationarity and exploration. In contrast, we study whether scalability can also be achieved via a disentangled representation. For this, we explicitly construct an object-centric intermediate representation to characterize the states of an environment, which we refer to as ‘semantic tracklets.’ We evaluate ‘semantic tracklets’ on the visual multi-agent particle environment (VMPE) and on the challenging visual multi-agent GFootball environment. ‘Semantic tracklets’ consistently outperform baselines on VMPE, and achieve a +2.4 higher score difference than baselines on GFootball. Notably, this method is the first to successfully learn a strategy for five players in the GFootball environment using only visual data. For more, please see our project page: https://ioujenliu.github.io/SemanticTracklets

The security of computers is at risk because of information leaking through their power consumption. Attackers can use advanced signal measurement and analysis to recover sensitive data from this side channel.To address this problem, this paper presents Maya, a simple and effective defense against power side channels. The idea is to use formal control to re-shape the power dissipated by a computer in an application-transparent manner—preventing attackers from learning any information about the applications that are running. With formal control, a controller can reliably keep power close to a desired target function even when runtime conditions change unpredictably. By selecting the target function intelligently, the controller can make power to follow any desired shape, appearing to carry activity information which, in reality, is unrelated to the application. Maya can be implemented in privileged software, firmware, or simple hardware. In this paper, we implement Maya on three machines using privileged threads only, and show its effectiveness and ease of deployment. Maya has already thwarted a newly-developed remote power attack.

Variational autoencoders (VAEs) are one of the powerful likelihood-based generative models with applications in many domains. However, they struggle to generate high-quality images, especially when samples are obtained from the prior without any tempering. One explanation for VAEs' poor generative quality is the prior hole problem: the prior distribution fails to match the aggregate approximate posterior. Due to this mismatch, there exist areas in the latent space with high density under the prior that do not correspond to any encoded image. Samples from those areas are decoded to corrupted images. To tackle this issue, we propose an energy-based prior defined by the product of a base prior distribution and a reweighting factor, designed to bring the base closer to the aggregate posterior. We train the reweighting factor by noise contrastive estimation, and we generalize it to hierarchical VAEs with many latent variable groups. Our experiments confirm that the proposed noise contrastive priors improve the generative performance of state-of-the-art VAEs by a large margin on the MNIST, CIFAR-10, CelebA 64, and CelebA HQ 256 datasets. Our method is simple and can be applied to a wide variety of VAEs to improve the expressivity of their prior distribution.

Modern approaches typically formulate semantic segmentation as a per-pixel classification task, while instance-level segmentation is handled with an alternative mask classification. Our key insight: mask classification is sufficiently general to solve both semantic- and instance-level segmentation tasks in a unified manner using the exact same model, loss, and training procedure. Following this observation, we propose MaskFormer, a simple mask classification model which predicts a set of binary masks, each associated with a single global class label prediction. Overall, the proposed mask classification-based method simplifies the landscape of effective approaches to semantic and panoptic segmentation tasks and shows excellent empirical results. In particular, we observe that MaskFormer outperforms per-pixel classification baselines when the number of classes is large. Our mask classification-based method outperforms both current state-of-the-art semantic (55.6 mIoU on ADE20K) and panoptic segmentation (52.7 PQ on COCO) models.

Machine learning advances in the last decade have relied significantly on large-scale datasets that continue to grow in size. Increasingly, those datasets also contain different data modalities. However, large multi-modal datasets are hard to annotate, and annotations may contain biases that we are often unaware of. Deep-net-based classifiers, in turn, are prone to exploit those biases and to find shortcuts. To study and quantify this concern, we introduce the perceptual score, a metric that assesses the degree to which a model relies on the different subsets of the input features, i.e., modalities. Using the perceptual score, we find a surprisingly consistent trend across four popular datasets: recent, more accurate state-of-the-art multi-modal models for visual question-answering or visual dialog tend to perceive the visual data less than their predecessors. This is concerning as answers are hence increasingly inferred from textual cues only. Using the perceptual score also helps to analyze model biases by decomposing the score into data subset contributions. We hope to spur a discussion on the perceptiveness of multi-modal models and also hope to encourage the community working on multi-modal classifiers to start quantifying perceptiveness via the proposed perceptual score.

We introduce REDO, a class-agnostic framework to REconstruct the Dynamic Objects from RGBD or calibrated videos. Compared to prior work, our problem setting is more realistic yet more challenging for three reasons: 1) due to occlusion or camera settings an object of interest may never be entirely visible, but we aim to reconstruct the complete shape; 2) we aim to handle different object dynamics including rigid motion, non-rigid motion, and articulation; 3) we aim to reconstruct different categories of objects with one unified framework. To address these challenges, we develop two novel modules. First, we introduce a canonical 4D implicit function which is pixel-aligned with aggregated temporal visual cues. Second, we develop a 4D transformation module which captures object dynamics to support temporal propagation and aggregation. We study the efficacy of REDO in extensive experiments on synthetic RGBD video datasets SAIL-VOS 3D and DeformingThings4D++, and on real-world video data 3DPW. We find REDO outperforms state-of-the-art dynamic reconstruction methods by a margin. In ablation studies we validate each developed component.

In practice, imitation learning is preferred over pure reinforcement learning whenever it is possible to design a teaching agent to provide expert supervision. However, we show that when the teaching agent makes decisions with access to privileged information that is unavailable to the student, this information is marginalized during imitation learning, resulting in an "imitation gap" and, potentially, poor results. Prior work bridges this gap via a progression from imitation learning to reinforcement learning. While often successful, gradual progression fails for tasks that require frequent switches between exploration and memorization. To better address these tasks and alleviate the imitation gap we propose 'Adaptive Insubordination' (ADVISOR). ADVISOR dynamically weights imitation and reward-based reinforcement learning losses during training, enabling on-the-fly switching between imitation and exploration. On a suite of challenging tasks set within gridworlds, multi-agent particle environments, and high-fidelity 3D simulators, we show that on-the-fly switching with ADVISOR outperforms pure imitation, pure reinforcement learning, as well as their sequential and parallel combinations.

The recent growth of web video sharing platforms has increased the demand for systems that can efficiently browse, retrieve and summarize video content. Query-aware multi-video summarization is a promising technique that caters to this demand. In this work, we introduce a novel Query-Aware Hierarchical Pointer Network for Multi-Video Summarization, termed DeepQAMVS, that jointly optimizes multiple criteria: (1) conciseness, (2) representativeness of important query-relevant events and (3) chronological soundness. We design a hierarchical attention model that factorizes over three distributions, each collecting evidence from a different modality, followed by a pointer network that selects frames to include in the summary. DeepQAMVS is trained with reinforcement learning, incorporating rewards that capture representativeness, diversity, query-adaptability and temporal coherence. We achieve state-of-the-art results on the MVS1K dataset, with inference time scaling linearly with the number of input video frames.

The success of deep learning in visual recognition tasks has driven advancements in multiple fields of research. Particularly, increasing attention has been drawn towards its application in agriculture. Nevertheless, while visual pattern recognition on farmlands carries enormous economic values, little progress has been made to merge computer vision and crop sciences due to the lack of suitable agricultural image datasets. Meanwhile, problems in agriculture also pose new challenges in computer vision. For example, semantic segmentation of aerial farmland images requires inference over extremely large-size images with extreme annotation sparsity. These challenges are not present in most of the common object datasets, and we show that they are more challenging than many other aerial image datasets. To encourage research in computer vision for agriculture, we present Agriculture-Vision: a large-scale aerial farmland image dataset for semantic segmentation of agricultural patterns. We collected 94,986 high-quality aerial images from 3,432 farmlands across the US, where each image consists of RGB and Near-infrared (NIR) channels with resolution as high as 10 cm per pixel. We annotate nine types of field anomaly patterns that are most important to farmers. As a pilot study of aerial agricultural semantic segmentation, we perform comprehensive experiments using popular semantic segmentation models; we also propose an effective model designed for aerial agricultural pattern recognition. Our experiments demonstrate several challenges Agriculture-Vision poses to both the computer vision and agriculture communities. Future versions of this dataset will include even more aerial images, anomaly patterns and image channels.

Understanding interactions between entities, e.g., joints of the human body, team sports players, etc., is crucial for tasks like forecasting. However, interactions between entities are commonly not observed and often hard to quantify. To address this challenge, recently, `Neural Relational Inference' was introduced. It predicts static relations between entities in a system and provides an interpretable representation of the underlying system dynamics that are used for better trajectory forecasting. However, generally, relations between entities change as time progresses. Hence, static relations improperly model the data. In response to this, we develop Dynamic Neural Relational Inference (dNRI), which incorporates insights from sequential latent variable models to predict separate relation graphs for every time-step. We demonstrate on several real-world datasets that modeling dynamic relations improves forecasting of complex trajectories.

Combinatorial optimization is frequently used in computer vision. For instance, in applications like semantic segmentation, human pose estimation and action recognition, programs are formulated for solving inference in Conditional Random Fields (CRFs) to produce a structured output that is consistent with visual features of the image. However, solving inference in CRFs is in general intractable, and approximation methods are computationally demanding and limited to unary, pairwise and hand-crafted forms of higher order potentials. In this paper, we show that we can learn program heuristics, i.e., policies, for solving inference in higher order CRFs for the task of semantic segmentation, using reinforcement learning. Our method solves inference tasks efficiently without imposing any constraints on the form of the potentials. We show compelling results on the Pascal VOC and MOTS datasets.

Weakly supervised learning has emerged as a compelling tool for object detection by reducing the need for strong supervision during training. However, major challenges remain: (1) differentiation of object instances can be ambiguous; (2) detectors tend to focus on discriminative parts rather than entire objects; (3) without ground truth, object proposals have to be redundant for high recalls, causing significant memory consumption. Addressing these challenges is difficult, as it often requires to eliminate uncertainties and trivial solutions. To target these issues we develop an instance-aware and context-focused unified framework. It employs an instance-aware self-training algorithm and a learnable Concrete DropBlock while devising a memory-efficient sequential batch back-propagation. Our proposed method achieves state-of-the-art results on COCO (12.1% AP, 24.8% AP50), VOC 2007 (54.9% AP), and VOC 2012 (52.1% AP), improving baselines by great margins. In addition, the proposed method is the first to benchmark ResNet based models and weakly supervised video object detection. Refer to our project page for code, models, and more details: https://github.com/NVlabs/wetectron.

In many vision-based reinforcement learning (RL) problems, the agent controls a movable object in its visual field, e.g., the player's avatar in video games and the robotic arm in visual grasping and manipulation. Leveraging action-conditioned video prediction, we propose an end-to-end learning frame-work to disentangle the controllable object from the observation signal. The disentangled representation is shown to be useful for RL as additional observation channels to the agent. Experiments on a set of Atari games with the popular Double DQN algorithm demonstrate improved sample efficiency and game performance (from 222.8% to 261.4% measured in normalized game scores, with prediction bonus reward).

Understanding of GAN training is still very limited. One major challenge is its non-convex-non-concave min-max objective, which may lead to sub-optimal local minima. In this work, we perform a global landscape analysis of the empirical loss of GANs. We prove that a class of separable-GAN, including the original JS-GAN, has exponentially many bad basins which are perceived as mode-collapse. We also study the relativistic pairing GAN (RpGAN) loss which couples the generated samples and the true samples. We prove that RpGAN has no bad basins. Experiments on synthetic data show that the predicted bad basin can indeed appear in training. We also perform experiments to support our theory that RpGAN has a better landscape than separable-GAN. For instance, we empirically show that RpGAN performs better than separable-GAN with relatively narrow neural nets. The code is available at \url{https://github.com/AilsaF/RS-GAN}.

Many recent datasets contain a variety of different data modalities, for instance, image, question, and answer data in visual question answering (VQA). When training deep net classifiers on those multi-modal datasets, the modalities get exploited at different scales, i.e., some modalities can more easily contribute to the classification results than others. This is suboptimal because the classifier is inherently biased towards a subset of the modalities. To alleviate this shortcoming, we propose a novel regularization term based on the functional entropy. Intuitively, this term encourages to balance the contribution of each modality to the classification result. However, regularization with the functional entropy is challenging. To address this, we develop a method based on the log-Sobolev inequality, which bounds the functional entropy with the functional-Fisher-information. Intuitively, this maximizes the amount of information that the modalities contribute. On the two challenging multi-modal datasets VQA-CPv2, and SocialIQ, we obtain state-of-the-art results while more uniformly exploiting the modalities. In addition, we demonstrate the efficacy of our method on Colored MNIST.

Various parallel actor-learner methods reduce long training times for deep reinforcement learning. Synchronous methods enjoy training stability while having lower data throughput. In contrast, asynchronous methods achieve high throughput but suffer from stability issues and lower sample efficiency due to ‘stale policies.’ To combine the advantages of both methods we propose High-Throughput Synchronous Deep Reinforcement Learning (HTS-RL). In HTS-RL, we perform learning and rollouts concurrently, devise a system design which avoids ‘stale policies’ and ensure that actors interact with environment replicas in an asynchronous manner while maintaining full determinism. We evaluate our approach on Atari games and the Google Research Football environment. Compared to synchronous baselines, HTS-RL is 2−6X faster. Compared to state-of-the-art asynchronous methods, HTS-RL has competitive throughput and consistently achieves higher average episode rewards.

Existing semi-supervised learning (SSL) algorithms use a single weight to balance the loss of labeled and unlabeled examples, i.e., all unlabeled examples are equally weighted. But not all unlabeled data are equal. In this paper we study how to use a different weight for “every” unlabeled example. Manual tuning of all those weights -- as done in prior work -- is no longer possible. Instead, we adjust those weights via an algorithm based on the influence function, a measure of a model's dependency on one training example. To make the approach efficient, we propose a fast and effective approximation of the influence function. We demonstrate that this technique outperforms state-of-the-art methods on semi-supervised image and language classification tasks.

Image captioning is an ambiguous problem, with many suitable captions for an image. To address ambiguity, beam search is the de facto method for sampling multiple captions. However, beam search is computationally expensive and known to produce generic captions. To address this concern, some variational auto-encoder (VAE) and generative adversarial net (GAN) based methods have been proposed. Though diverse, GAN and VAE are less accurate. In this paper, we first predict a meaningful summary of the image, then generate the caption based on that summary. We use part-of-speech as summaries, since our summary should drive caption generation. We achieve the trifecta: (1) High accuracy for the diverse captions as evaluated by standard captioning metrics and user studies; (2) Faster computation of diverse captions compared to beam search and diverse beam search; and (3) High diversity as evaluated by counting novel sentences, distinct n-grams and mutual overlap (i.e., mBleu-4) scores.

Generative adversarial nets (GANs) and variational auto-encoders have significantly improved our distribution modeling capabilities, showing promise for dataset augmentation, image-to-image translation and feature learning. However, to model high-dimensional distributions, sequential training and stacked architectures are common, increasing the number of tunable hyper-parameters as well as the training time. Nonetheless, the sample complexity of the distance metrics remains one of the factors affecting GAN training. We first show that the recently proposed sliced Wasserstein distance has compelling sample complexity properties when compared to the Wasserstein distance. To further improve the sliced Wasserstein distance we then analyze its `projection complexity' and develop the max-sliced Wasserstein distance which enjoys compelling sample complexity while reducing projection complexity, albeit necessitating a max estimation. We finally illustrate that the proposed distance trains GANs on high-dimensional images up to a resolution of 256x256 easily.

We introduce SAIL-VOS (Semantic Amodal Instance Level Video Object Segmentation), a new dataset aiming to stimulate semantic amodal segmentation research. Humans can effortlessly recognize partially occluded objects and reliably estimate their spatial extent beyond the visible. However, few modern computer vision techniques are capable of reasoning about occluded parts of an object. This is partly due to the fact that very few image datasets and no video dataset exist which permit development of those methods. To address this issue, we present a synthetic dataset extracted from the photo-realistic game GTA-V. Each frame is accompanied with densely annotated, pixel-accurate visible and amodal segmentation masks with semantic labels. More than 1.8M objects are annotated resulting in 100 times more annotations than existing datasets. We demonstrate the challenges of the dataset by quantifying the performance of several baselines. Data and additional material is available at http://sailvos.web.illinois.edu.

Collaboration is a necessary skill to perform tasks that are beyond one agent's capabilities. Addressed extensively in both conventional and modern AI, multi-agent collaboration has often been studied in the context of simple grid worlds. We argue that there are inherently visual aspects to collaboration which should be studied in visually rich environments. A key element in collaboration is communication that can be either explicit, through messages, or implicit, through perception of the other agents and the visual world. Learning to collaborate in a visual environment entails learning (1) to perform the task, (2) when and what to communicate, and (3) how to act based on these communications and the perception of the visual world. In this paper we study the problem of learning to collaborate directly from pixels in AI2-THOR and demonstrate the benefits of explicit and implicit modes of communication to perform visual tasks. Refer to our project page for more details: https://prior.allenai.org/projects/two-body-problem.

The recently proposed audio-visual scene-aware dialog task paves the way to a more data-driven way of learning virtual assistants, smart speakers and car navigation systems. However, very little is known to date about how to effectively extract meaningful information from a plethora of sensors that pound the computational engine of those devices. Therefore, in this paper, we provide and carefully analyze a simple baseline for audio-visual scene-aware dialog which is trained end-to-end. Our method differentiates in a data-driven manner useful signals from distracting ones using an attention mechanism. We evaluate the proposed approach on the recently introduced and challenging audio-visual scene-aware dataset, and demonstrate the key features that permit to outperform the current state-of-the-art by more than 20% on CIDEr.

Dialog is an effective way to exchange information, but subtle details and nuances are extremely important. While significant progress has paved a path to address visual dialog with algorithms, details and nuances remain a challenge. Attention mechanisms have demonstrated compelling results to extract details in visual question answering and also provide a convincing framework for visual dialog due to their interpretability and effectiveness. However, the many data utilities that accompany visual dialog challenge existing attention techniques. We address this issue and develop a general attention mechanism for visual dialog which operates on any number of data utilities. To this end, we design a factor graph based attention mechanism which combines any number of utility representations. We illustrate the applicability of the proposed approach on the challenging and recently introduced VisDial datasets, outperforming recent state-of-the-art methods by 1.1% for VisDial0.9 and by 2% for VisDial1.0 on MRR. Our ensemble model improved the MRR score on VisDial1.0 by more than 6%.

In this paper, we propose a new generative model for multi-agent trajectory data, focusing on the case of multi-player sports games. Our model leverages graph neural networks (GNNs) and variational recurrent neural networks (VRNNs) to achieve a permutation equivariant model suitable for sports. On two challenging datasets (basketball and soccer), we show that we are able to produce more accurate forecasts than previous methods. We assess accuracy using various metrics, such as log-likelihood and "best of N" loss, based on N different samples of the future. We also measure the distribution of statistics of interest, such as player location or velocity, and show that the distribution induced by our generative model better matches the empirical distribution of the test set. Finally, we show that our model can perform conditional prediction, which lets us answer counterfactual questions such as “how will the players move differently if A passes the ball to B instead of C?”

Diverse and accurate vision+language modeling is an important goal to retain creative freedom and maintain user engagement. However, adequately capturing the intricacies of diversity in language models is challenging. Recent works commonly resort to latent variable models augmented with more or less supervision from object detectors or part-of-speech tags. In common to all those methods is the fact that the latent variable either only initializes the sentence generation process or is identical across the steps of generation. Both methods offer no fine-grained control. To address this concern, we propose Seq-CVAE which learns a latent space for every word. We encourage this temporal latent space to capture the 'intention' about how to complete the sentence by mimicking a representation which summarizes the future. We illustrate the efficacy of the proposed approach on the challenging MSCOCO dataset, significantly improving diversity metrics compared to baselines while performing on par w.r.t. sentence quality.

We propose to learn word embeddings from visual co-occurrences. Two words co-occur visually if both words apply to the same image or image region. Specifically, we extract four types of visual co-occurrences between object and attribute words from large-scale, textually-annotated visual databases like VisualGenome and ImageNet. We then train a multi-task log-bilinear model that compactly encodes word “meanings” represented by each co-occurrence type into a single visual word-vector. Through unsupervised clustering, supervised partitioning, and a zero-shot-like generalization analysis we show that our word embeddings complement text-only embeddings like Glove by better representing similarities and differences between visual concepts that are difficult to obtain from text corpora alone. We further evaluate our embeddings on five downstream applications, four of which are vision-language tasks. Augmenting Glove with our embeddings yields gains on all tasks. We also find that random embeddings perform comparably to learned embeddings on all supervised vision-language tasks, contrary to conventional wisdom.

We show that for human-object interaction detection a relatively simple factorized model with appearance and layout encodings constructed from pre-trained object detectors outperforms more sophisticated approaches. Our model includes factors for detection scores, human and object appearance, and coarse (box-pair configuration) and optionally fine-grained layout (human pose). We also develop training techniques that improve learning efficiency by: (1) eliminating a train-inference mismatch; (2) rejecting easy negatives during mini-batch training; and (3) using a ratio of negatives to positives that is two orders of magnitude larger than existing approaches. We conduct a thorough ablation study to understand the importance of different factors and training techniques using the challenging HICO-Det dataset.

Reinforcement learning (RL) has attracted much attention recently, as new and emerging AI-based applications are demanding the capabilities to intelligently react to environment changes. Unlike distributed deep neural network (DNN) training, the distributed RL training has its unique workload characteristics - it generates orders of magnitude more iterations with much smaller sized but more frequent gradient aggregations. More specifically, our study with typical RL algorithms shows that their distributed training is latency critical and that the network communication for gradient aggregation occupies up to 83.2% of the execution time of each training iteration.

Inferring the most likely configuration for a subset of variables of a joint distribution given the remaining ones -- which we refer to as co-generation -- is an important challenge that is computationally demanding for all but the simplest settings. This task has received a considerable amount of attention, particularly for classical ways of modeling distributions like structured prediction. In contrast, almost nothing is known about this task when considering recently proposed techniques for modeling high-dimensional distributions, particularly generative adversarial nets (GANs). Therefore, in this paper, we study the occurring challenges for co-generation with GANs. To address those challenges we develop an annealed importance sampling based Hamiltonian Monte Carlo co-generation algorithm. The presented approach significantly outperforms classical gradient based methods on a synthetic and on the CelebA and LSUN datasets.

For joint inference over multiple variables, a variety of structured prediction techniques have been developed to model correlations among variables and thereby improve predictions. However, many classical approaches suffer from one of two primary drawbacks: they either lack the ability to model high-order correlations among variables while maintaining computationally tractable inference, or they do not allow to explicitly model known correlations. To address this shortcoming, we introduce ‘Graph Structured Prediction Energy Networks,’ for which we develop inference techniques that allow to both model explicit local and implicit higher-order correlations while maintaining tractability of inference. We apply the proposed method to tasks from the natural language processing and computer vision domain and demonstrate its general utility.

Reasoning is an important ability that we learn from a very early age. Yet, reasoning is extremely hard for algorithms. Despite impressive recent progress that has been reported on tasks that necessitate reasoning, such as visual question answering and visual dialog, models often exploit biases in datasets. To develop models with better reasoning abilities, recently, the new visual commonsense reasoning(VCR) task has been introduced. Not only do models have to answer questions, but also do they have to provide a reason for the given answer. The proposed baseline achieved compelling results, leveraging a meticulously designed model composed of LSTM modules and attention nets. Here we show that a much simpler model obtained by ablating and pruning the existing intricate baseline can perform better with half the number of trainable parameters. By associating visual features with attribute information and better text to image grounding, we obtain further improvements for our simpler & effective baseline, TAB-VCR. We show that this approach results in a 5.3%, 4.4% and 6.5% absolute improvement over the previous state-of-the-art on question answering, answer justification and holistic VCR. Webpage: https://deanplayerljx.github.io/tabvcr/

We propose Chirality Nets, a family of deep nets that is equivariant to the “chirality transform,” i.e., the transformation to create a chiral pair. Through parameter sharing, odd and even symmetry, we propose and prove variants of standard building blocks of deep nets that satisfy the equivariance property, including fully connected layers, convolutional layers, batch-normalization, and LSTM/GRU cells. The proposed layers lead to a more data efficient representation and a reduction in computation by exploiting symmetry. We evaluate chirality nets on the task of human pose regression, which naturally exploits the left/right mirroring of the human body. We study three pose regression tasks: 3D pose estimation from video, 2D pose forecasting, and skeleton based activity recognition. Our approach achieves/matches state-of-the-art results, with more significant gains on small datasets and limited-data settings.

Image captioning is an important task, applicable to virtual assistants, editing tools, image indexing, and support of the disabled. In recent years significant progress has been made in image captioning, using Recurrent Neural Networks powered by long-short-term-memory (LSTM) units. Despite mitigating the vanishing gradient problem, and despite their compelling ability to memorize dependencies, LSTM units are complex and inherently sequential across time. To address this issue, recent work has shown benefits of convolutional networks for machine translation and conditional image generation [9, 34, 35]. Inspired by their success, in this paper, we develop a convolutional image captioning technique. We demonstrate its efficacy on the challenging MSCOCO dataset and demonstrate performance on par with the LSTM baseline [16], while having a faster training time per number of parameters. We also perform a detailed analysis, providing compelling reasons in favor of convolutional language generation approaches.

Generative Adversarial Nets (GANs) are very successful at modeling distributions from given samples, even in the high-dimensional case. However, their formulation is also known to be hard to optimize and often not stable. While this is particularly true for early GAN formulations, there has been significant empirically motivated and theoretically founded progress to improve stability, for instance, by using the Wasserstein distance rather than the Jenson-Shannon divergence. Here, we consider an alternative formulation for generative modeling based on random projections which, in its simplest form, results in a single objective rather than a saddle-point formulation. By augmenting this approach with a discriminator we improve its accuracy. We found our approach to be significantly more stable compared to even the improved Wasserstein GAN. Further, unlike the traditional GAN loss, the loss formulated in our method is a good measure of the actual distance between the distributions and, for the first time for GAN training, we are able to show estimates for the same.

Human conversation is a complex mechanism with subtle nuances. It is hence an ambitious goal to develop artificial intelligence agents that can participate fluently in a conversation. While we are still far from achieving this goal, recent progress in visual question answering, image captioning, and visual question generation shows that dialog systems may be realizable in the not too distant future. To this end, a novel dataset was introduced recently and encouraging results were demonstrated, particularly for question answering. In this paper, we demonstrate a simple symmetric discriminative baseline, that can be applied to both predicting an answer as well as predicting a question. We show that this method performs on par with the state of the art, even memory net based methods. In addition, for the first time on the visual dialog dataset, we assess the performance of a system asking questions, and demonstrate how visual dialog can be generated from discriminative question generation and question answering.

Textual grounding, i.e., linking words to objects in images, is a challenging but important task for robotics and human-computer interaction. Existing techniques benefit from recent progress in deep learning and generally formulate the task as a supervised learning problem, selecting a bounding box from a set of possible options. To train these deep net based approaches, access to a large-scale datasets is required, however, constructing such a dataset is time-consuming and expensive. Therefore, we develop a completely unsupervised mechanism for textual grounding using hypothesis testing as a mechanism to link words to detected image concepts. We demonstrate our approach on the ReferIt Game dataset and the Flickr30k data, outperforming baselines by 7.98% and 6.96% respectively.

Many image restoration problems are ill-posed in nature, hence, beyond the input image, most existing methods rely on a carefully engineered image prior, which enforces some local image consistency in the recovered image. How tightly the prior assumptions are fulfilled has a big impact on the resulting task performance. To obtain more flexibility, in this work, we proposed to design the image prior in a data-driven manner. Instead of explicitly defining the prior, we learn it using deep generative models. We demonstrate that this learned prior can be applied to many image restoration problems using an unified framework.

Training real-world Deep Neural Networks (DNNs) can take an eon (i.e., weeks or months) without leveraging distributed systems. Even distributed training takes inordinate time, of which a large fraction is spent in communicating weights and gradients over the network. State-of-the-art distributed training algorithms use a hierarchy of worker-aggregator nodes. The aggregators repeatedly receive gradient updates from their allocated group of the workers, and send back the updated weights. This paper sets out to reduce this significant communication cost by embedding data compression accelerators in the Network Interface Cards (NICs). To maximize the benefits of in-network acceleration, the proposed solution, named INCEPTIONN (In-Network Computing to Exchange and Process Training Information Of Neural Networks), uniquely combines hardware and algorithmic innovations by exploiting the following three observations. (1) Gradients are significantly more tolerant to precision loss than weights and as such lend themselves better to aggressive compression without the need for the complex mechanisms to avert any loss. (2) The existing training algorithms only communicate gradients in one leg of the communication, which reduces the opportunities for in-network acceleration of compression. (3) The aggregators can become a bottleneck with compression as they need to compress/decompress multiple streams from their allocated worker group. To this end, we first propose a lightweight and hardware-friendly lossy-compression algorithm for floating-point gradients, which exploits their unique value characteristics. This compression not only enables significantly reducing the gradient communication with practically no loss of accuracy, but also comes with low complexity for direct implementation as a hardware block in the NIC. To maximize the opportunities for compression and avoid the bottleneck at aggregators, we also propose an aggregator-free training algorithm that exchanges gradients in both legs of communication in the group, while the workers collectively perform the aggregation in a distributed manner. Without changing the mathematics of training, this algorithm leverages the associative property of the aggregation operator and enables our in-network accelerators to (1) apply compression for all communications, and (2) prevent the aggregator nodes from becoming bottlenecks. Our experiments demonstrate that INCEPTIONN reduces the communication time by 70.9~80.7% and offers 2.2~3.1x speedup over the conventional training system, while achieving the same level of accuracy.

Deep structured models are widely used for tasks like semantic segmentation, where explicit correlations between variables provide important prior information which generally helps to reduce the data needs of deep nets. However, current deep structured models are restricted by oftentimes very local neighborhood structure, which cannot be increased for computational complexity reasons, and by the fact that the output configuration, or a representation thereof, cannot be transformed further. Very recent approaches which address those issues include graphical model inference inside deep nets so as to permit subsequent non-linear output space transformations. However, optimization of those formulations is challenging and not well understood. Here, we develop a novel model which generalizes existing approaches, such as structured prediction energy networks, and discuss a formulation which maintains applicability of existing inference techniques.

Distributed training of deep nets is an important technique to address some of the present day computing challenges like memory consumption and computational demands. Classical distributed approaches, synchronous or asynchronous, are based on the parameter server architecture, i.e., worker nodes compute gradients which are communicated to the parameter server while updated parameters are returned. Recently, distributed training with AllReduce operations gained popularity as well. While many of those operations seem appealing, little is reported about wall-clock training time improvements. In this paper, we carefully analyze the AllReduce based setup, propose timing models which include network latency, bandwidth, cluster size and compute time, and demonstrate that a pipelined training with a width of two combines the best of both synchronous and asynchronous training. Specifically, for a setup consisting of a four-node GPU cluster we show wall-clock time training improvements of up to 5.4x compared to conventional approaches.

Accurately answering a question about a given image requires combining observations with general knowledge. While this is effortless for humans, reasoning with general knowledge remains an algorithmic challenge. To advance research in this direction a novel fact-based' visual question answering (FVQA) task has been introduced recently along with a large set of curated facts which link two entities, i.e., two possible answers, via a relation. Given a question-image pair, deep network techniques have been employed to successively reduce the large set of facts until one of the two entities of the final remaining fact is predicted as the answer. We observe that a successive process which considers one fact at a time to form a local decision is sub-optimal. Instead, we develop an entity graph and use a graph convolutional network toreason' about the correct answer by jointly considering all entities. We show on the challenging FVQA dataset that this leads to an improvement in accuracy of around 7% compared to the state-of-the-art.

Data parallelism can boost the training speed of convolutional neural networks (CNN), but could suffer from significant communication costs caused by gradient aggregation. To alleviate this problem, several scalar quantization techniques have been developed to compress the gradients. But these techniques could perform poorly when used together with decentralized aggregation protocols like ring all-reduce (RAR), mainly due to their inability to directly aggregate compressed gradients. In this paper, we empirically demonstrate the strong linear correlations between CNN gradients, and propose a gradient vector quantization technique, named GradiVeQ, to exploit these correlations through principal component analysis (PCA) for substantial gradient dimension reduction. GradiveQ enables direct aggregation of compressed gradients, hence allows us to build a distributed learning system that parallelizes GradiveQ gradient compression and RAR communications. Extensive experiments on popular CNNs demonstrate that applying GradiveQ slashes the wall-clock gradient aggregation time of the original RAR by more than 5x without noticeable accuracy loss, and reduce the end-to-end training time by almost 50%. The results also show that \GradiveQ is compatible with scalar quantization techniques such as QSGD (Quantized SGD), and achieves a much higher speed-up gain under the same compression ratio.

Generating diverse questions for given images is an important task for computational education, entertainment and AI assistants. Different from many conventional prediction techniques is the need for algorithms to generate a diverse set of plausible questions, which we refer to as creativity. In this paper we propose a creative algorithm for visual question generation which combines the advantages of variational autoencoders with long short-term memory networks. We demonstrate that our framework is able to generate a large set of varying questions given a single input image.

Semantic image inpainting is a challenging task where large missing regions have to be filled based on the available visual data. Existing methods which extract information from only a single image generally produce unsatisfactory results due to the lack of high level context. In this paper, we propose a novel method for semantic image inpainting, which generates the missing content by conditioning on the available data. Given a trained generative model, we search for the closest encoding of the corrupted image in the latent image manifold using our context and prior losses. This encoding is then passed through the generative model to infer the missing content. In our method, inference is possible irrespective of how the missing content is structured, while the state-of-the-art learning based method requires specific information about the holes in the training phase. Experiments on three datasets show that our method successfully predicts information in large missing regions and achieves pixel-level photorealism, significantly outperforming the state-of-the-art methods.

This paper explores image caption generation using conditional variational auto-encoders (CVAEs). Standard CVAEs with a fixed Gaussian prior yield descriptions with too little variability. Instead, we propose two models that explicitly structure the latent space around K components corresponding to different types of image content, and combine components to create priors for images that contain multiple types of content simultaneously (e.g., several kinds of objects). Our first model uses a Gaussian Mixture model (GMM) prior, while the second one defines a novel Additive Gaussian (AG) prior that linearly combines component means. We show that both models produce captions that are more diverse and more accurate than a strong LSTM baseline or a “vanilla” CVAE with a fixed Gaussian prior, with AG-CVAE showing particular promise.

Instance level video object segmentation is an important technique for video editing and compression. To capture the temporal coherence, in this paper, we develop MaskRNN, a recurrent neural net approach which fuses in each frame the output of two deep nets for each object instance - a binary segmentation net providing a mask and a localization net providing a bounding box. Due to the recurrent component and the localization component, our method is able to take advantage of long-term temporal structures of the video data as well as rejecting outliers. We validate the proposed algorithm on three challenging benchmark datasets, the DAVIS-2016 dataset, the DAVIS-2017 dataset, and the Segtrack v2 dataset, achieving state-of-the-art performance on all of them.

Generative adversarial nets (GANs) are a promising technique for modeling a distribution from samples. It is however well known that GAN training suffers from instability due to the nature of its saddle point formulation. In this paper, we explore ways to tackle the instability problem by dualizing the discriminator. We start from linear discriminators in which case conjugate duality provides a mechanism to reformulate the saddle point objective into a maximization problem, such that both the generator and the discriminator of this ‘dualing GAN’ act in concert. We then demonstrate how to extend this intuition to non-linear formulations. For GANs with linear discriminators our approach is able to remove the instability in training, while for GANs with nonlinear discriminators our approach provides an alternative to the commonly used GAN training algorithm.

Finding the maximum a-posteriori (MAP) assignment is a central task in graphical models. Since modern applications give rise to very large problem instances, there is increasing need for efficient solvers. In this work we propose to improve the efficiency of coordinate-minimization-based dual-decomposition solvers by running their updates asynchronously in parallel. In this case message-passing inference is performed by multiple processing units simultaneously without coordination, all reading and writing to shared memory. We analyze the convergence properties of the resulting algorithms and identify settings where speedup gains can be expected. Our numerical evaluations show that this approach indeed achieves significant speedups in common computer vision tasks.

The quest for algorithms that enable cognitive abilities is an important part of machine learning. A common trait in many recently investigated cognitive-like tasks is that they take into account different data modalities, such as visual and textual input. In this paper we propose a novel and generally applicable form of attention mechanism that learns high-order correlations between various data modalities. We show that high-order correlations effectively direct the appropriate attention to the relevant elements in the different data modalities that are required to solve the joint task. We demonstrate the effectiveness of our high-order attention mechanism on the task of visual question answering (VQA), where we achieve state-of-the-art performance on the standard VQA dataset.

Textual grounding is an important but challenging task for human-computer inter- action, robotics and knowledge mining. Existing algorithms generally formulate the task as selection from a set of bounding box proposals obtained from deep net based systems. In this work, we demonstrate that we can cast the problem of textual grounding into a unified framework that permits efficient search over all possible bounding boxes. Hence, the method is able to consider significantly more proposals and doesn’t rely on a successful first stage hypothesizing bounding box proposals. Beyond, we demonstrate that the trained parameters of our model can be used as word-embeddings which capture spatial-image relationships and provide interpretability. Lastly, at the time of submission, our approach outperformed the current state-of-the-art methods on the Flickr 30k Entities and the ReferItGame dataset by 3.08% and 7.77% respectively.

In the past year, convolutional neural networks have been shown to perform extremely well for stereo estimation. However, current architectures rely on siamese networks which exploit concatenation followed by further processing layers, requiring a minute of GPU computation per image pair. In contrast, in this paper we propose a matching network which is able to produce very accurate results in less than a second of GPU computation. Towards this goal, we exploit a product layer which simply computes the inner product between the two representations of a siamese architecture. We train our network by treating the problem as multi-class classification, where the classes are all possible disparities. This allows us to get calibrated scores, which result in much better matching performance when compared to existing approaches.

Supervised training of deep neural nets typically relies on minimizing cross-entropy. However, in many domains, we are interested in performing well on metrics specific to the application. In this paper we propose a direct loss minimization approach to train deep neural networks, which provably minimizes the application-specific loss function. This is often non-trivial, since these functions are neither smooth nor decomposable and thus are not amenable to optimization with standard gradient-based methods. We demonstrate the effectiveness of our approach in the context of maximizing average precision for ranking problems. Towards this goal, we develop a novel dynamic programming algorithm that can efficiently compute the weight updates. Our approach proves superior to a variety of baselines in the context of action classification and object detection, especially in the presence of label noise.

In this paper we aim at facilitating generalization for deep networks while supporting interpretability of the learned representations. Towards this goal, we propose a clustering based regularization that encourages parsimonious representations. Our k-means style objective is easy to optimize and flexible supporting various forms of clustering, including sample and spatial clustering as well as co-clustering. We demonstrate the effectiveness of our approach on the tasks of unsupervised learning, classification, fine grained categorization and zero-shot learning.

Inference in Markov random fields subject to consistency structure is a fundamental problem that arises in many real-life applications. In order to enforce consistency, classical approaches utilize consistency potentials or encode constraints over feasible instances. Unfortunately this comes at the price of a serious computational bottleneck. In this paper we suggest to tackle consistency by incorporating constraints on beliefs. This permits derivation of a closed-form message-passing algorithm which we refer to as the Constraints Based Convex Belief Propagation (CBCBP). Experiments show that CBCBP outperforms the standard approach while being at least an order of magnitude faster.

Conditional random fields maximize the log-likelihood of training labels given the training data, e.g., objects given images. In many cases the training labels are structures that consist of a set of variables and the computational complexity for estimating their likelihood is exponential in the number of the variables. Learning algorithms relax this computational burden using approximate inference that is nested as a sub- procedure. In this paper we describe the objective function for nested learning and inference in conditional random fields. The devised objective maximizes the log-beliefs --- probability distributions over subsets of training variables that agree on their marginal probabilities. This objective is concave and consists of two types of variables that are related to the learning and inference tasks respectively. Importantly, we afterwards show how to blend the learning and inference procedure and effectively get to the identical optimum much faster. The proposed algorithm currently achieves the state-of- the-art in various computer vision applications.

The goal of this paper is to enable a 3D “virtual-tour” of an apartment given a small set of monocular images of different rooms, as well as a 2D floor plan. We frame the problem as inference in a Markov Random Field which reasons about the layout of each room and its relative pose (3D rotation and translation) within the full apartment. This gives us accurate camera pose in the apartment for each image. What sets us apart from past work in layout estimation is the use of floor plans as a source of prior knowledge, as well as localization of each image within a bigger space (apartment). In particular, we exploit the floor plan to impose aspect ratio constraints across the layouts of different rooms, as well as to extract semantic information, e.g., the location of windows which are marked in floor plans. We show that this information can significantly help in resolving the challenging room-apartment alignment problem. We also derive an efficient exact inference algorithm which takes only a few ms per apartment. This is due to the fact that we exploit integral geometry as well as our new bounds on the aspect ratio of rooms which allow us to carve the space, significantly reducing the number of physically possible configurations. We demonstrate the effectiveness of our approach on a new dataset which contains over 200 apartments.

Despite the promising performance of conventional fully supervised algorithms, semantic segmentation has remained an important, yet challenging task. Due to the limited availability of complete annotations, it is of great interest to design solutions for semantic segmentation that take into account weakly labeled data, which is readily available at a much larger scale. Contrasting the common theme to develop a different algorithm for each type of weak annotation, in this work, we propose a unified approach that incorporates various forms of weak supervision - image level tags, bounding boxes, and partial labels - to produce a pixel-wise labeling. We conduct a rigorous evaluation on the challenging Siftflow dataset for various weakly labeled settings, and show that our approach outperforms the state-of-the-art by 12% on per-class accuracy, while maintaining comparable per-pixel accuracy.

In this paper we tackle the problem of instance-level segmentation and depth ordering from a single monocular image. Towards this goal, we take advantage of convolutional neural nets and train them to directly predict instance-level segmentations where the instance ID encodes the depth ordering within image patches. To provide a coherent single explanation of an image we develop a Markov random field which takes as input the predictions of convolutional neural nets applied at overlapping patches of different resolutions, as well as the output of a connected component algorithm. It aims to predict accurate instance-level segmentation and depth ordering. We demonstrate the effectiveness of our approach on the challenging KITTI benchmark and show good performance on both tasks.

Many problems in real-world applications involve predicting several random variables that are statistically related. Markov random fields (MRFs) are a great mathematical tool to encode such dependencies. The goal of this paper is to combine MRFs with deep learning to estimate complex representations while taking into account the dependencies between the output random variables. Towards this goal, we propose a training algorithm that is able to learn structured models jointly with deep features that form the MRF potentials. Our approach is efficient as it blends learning and inference and makes use of GPU acceleration. We demonstrate the effectiveness of our algorithm in the tasks of predicting words from noisy images, as well as tagging of Flickr photographs. We show that joint learning of the deep features and the MRF parameters results in significant performance gains.

Maximum a-posteriori (MAP) inference is an important task for many applications. Although the standard formulation gives rise to a hard combinatorial optimization problem, several effective approximations have been proposed and studied in recent years. We focus on linear programming (LP) relaxations, which have achieved state-of-the-art performance in many applications. However, optimization of the resulting program is in general challenging due to non-smoothness and complex non-separable constraints.Therefore, in this work we study the benefits of augmenting the objective function of the relaxation with strong convexity. Specifically, we introduce strong convexity by adding a quadratic term to the LP relaxation objective. We provide theoretical guarantees for the resulting programs, bounding the difference between their optimal value and the original optimum. Further, we propose suitable optimization algorithms and analyze their convergence.

Key point matching in images of indoor scenes traditionally employs features like SIFT, GIST and HOG. While those features work very well for two images related to each other by small camera transformations, we commonly observe a drop in performance for patches representing scene elements visualized from a very different perspective. Since increasing the space of considered local transformations for feature matching decreases their discriminative abilities, we propose a more global approach inspired by the recent success of monocular scene understanding. In particular we propose to reconstruct a box-like model of the scene from every single image and use it to rectify images before matching. We show that a monocular scene model reconstruction and rectification preceding standard feature matching significantly improves key point matching and dramatically improves reconstruction of difficult indoor scenes.

Computational education offers an important add-on to conventional teaching. To provide optimal learning conditions, accurate representation of students’ current skills and adaptation to newly acquired knowledge are essential. To obtain sufficient representational power we investigate suitability of general graphical models and discuss adaptation by learning parameters of a log-linear distribution. For interpretability we propose to constrain the parameter space a-priori by leveraging domain knowledge. We show the benefits of general graphical models and of regularizing the parameter space by evaluation of our models on data collected from a computational education software for children having difficulties in learning mathematics.

We propose a sequential optimization technique for segmenting a rectified image of a facade into semantic categories. Our method retrieves a parsing which respects common architectural constraints and also returns a certificate for global optimality. Contrasting the suggested method, the considered facade labeling problem is typically tackled as a classification task or as grammar parsing. Both approaches are not capable of fully exploiting the regularity of the problem. Therefore, our technique very significantly improves the accuracy compared to the state-of-the-art while being an order of magnitude faster. In addition, in 85% of the test images we obtain a certificate for optimality.

We tackle the problem of weakly labeled semantic segmentation, where the only source of annotation are image tags encoding which classes are present in the scene. This is an extremely difficult problem as no pixel-wise labelings are available, not even at training time. In this paper, we show that this problem can be formalized as an instance of learning in a latent structured prediction framework, where the graphical model encodes the presence and absence of a class as well as the assignments of semantic labels to superpixels. As a consequence, we are able to leverage standard algorithms with good theoretical properties. We demonstrate the effectiveness of our approach using the challenging SIFT-flow dataset and show average per-class accuracy improvements of 7% over the state-of-the-art.

While MAP inference is typically intractable for many real-world applications, linear programming relaxations have been proven very effective. Dual block-coordinate descent methods are among the most efficient solvers, however, they are prone to get stuck in sub-optimal points. Although subgradient approaches achieve global convergence, they are typically slower in practice. To improve convergence speed, algorithms which compute the steepest ε-descent direction by solving a quadratic program have been proposed. In this paper we suggest to decouple the quadratic program based on the Frank-Wolfe approach. This allows us to obtain an efficient and easy to parallelize algorithm while retaining the global convergence properties. Our method proves superior when compared to existing algorithms on a set of spin-glass models and protein design tasks.

In this paper we propose an approach to jointly infer the room layout as well as the objects present in the scene. Towards this goal, we propose a branch and bound algorithm which is guaranteed to retrieve the global optimum of the joint problem. The main difficulty resides in taking into account occlusion in order to not over-count the evidence. We introduce a new decomposition method, which generalizes integral geometry to triangular shapes, and allows us to bound the different terms in constant time. We exploit both geometric cues and object detectors as image features and show large improvements in 2D and 3D object detection over state-of-the-art deformable part-based models.

In this paper we propose an approach to jointly estimate the layout of rooms as well as the clutter present in the scene using RGB-D data. Towards this goal, we propose an effective model that is able to exploit both depth and appearance features, which are complementary. Furthermore, our approach is efficient as we exploit the inherent decomposition of additive potentials. We demonstrate the effectiveness of our approach on the challenging NYU v2 dataset and show that employing depth reduces the layout error by 6% and the clutter estimation by 13%.

In this paper we present active learning algorithms in the context of structured prediction problems. To reduce the amount of labeling necessary to learn good models, our algorithms only label subsets of the output. To this end, we query examples using entropies of local marginals, which are a good surrogate for uncertainty. We demonstrate the effectiveness of our approach in the task of 3D layout prediction from single images, and show that good models are learned when labeling only a handful of random variables. In particular, the same performance as using the full training set can be obtained while only labeling ~10\% of the random variables.

Existing approaches to indoor scene understanding formulate the problem as a structured prediction task focusing on estimating the 3D bounding box which best describes the scene layout. Unfortunately, these approaches utilize high order potentials which are computationally intractable and rely on ad-hoc approximations for both learning and inference. In this paper we show that the potentials commonly used in the literature can be decomposed into pair-wise potentials by extending the concept of integral images to geometry. As a consequence no heuristic reduction of the search space is required. In practice, this results in large improvements in performance over the state-of-the-art, while being orders of magnitude faster.

While finding the exact solution for the MAP inference problem is intractable for many real-world tasks, MAP LP relaxations have been shown to be very effective in practice. However, the most efficient methods that perform block coordinate descent can get stuck in sub-optimal points as they are not globally convergent. In this work we propose to augment these algorithms with an $\epsilon$-descent approach and present a method to efficiently optimize for a descent direction in the subdifferential using a margin-based extension of the Fenchel-Young duality theorem. Furthermore, the presented approach provides a methodology to construct a primal optimal solution from its dual optimal counterpart. We demonstrate the efficiency of the presented approach on spin glass models and protein interactions problems and show that our approach outperforms state-of-the-art solvers.

In this paper we propose a distributed message-passing algorithm for inference in large scale graphical models. Our method can handle large problems efficiently by distributing and parallelizing the computation and memory requirements. The convergence and optimality guarantees of recently developed message-passing algorithms are preserved by introducing new types of consistency messages, sent between the distributed computers. We demonstrate the effectiveness of our approach in the task of stereo reconstruction from high-resolution imagery, and show that inference is possible with more than 200 labels in images larger than 10 MPixels.

How many people should you ask if you are not sure about your way? We provide an answer to this question for Random Forest classification. The presented method is based on the statistical formulation of confidence intervals and conjugate priors for binomial as well as multinomial distributions. We derive appealing decision rules to speed up the classification process by leveraging the fact that many samples can be clearly mapped to classes. Results on test data are provided, and we highlight the applicability of our method to a wide range of problems. The approach introduces only one non-heuristic parameter, that allows to trade-off accuracy and speed without any re-training of the classifier. The proposed method automatically adapts to the difficulty of the test data and makes classification significantly faster without deteriorating the accuracy.

For the single-group multicast scenario, where K users are served with the same data by a base station equipped with N antennas, we present two beamforming algorithms which outperform state-of-the-art multicast filters and feature a drastically reduced complexity at the same time. For the power minimization problem, where QoS constraints need to be satisfied, we introduce a successive beamforming filter computation approach aiming at satisfying at least one additional SNR constraint per orthogonal filter update. As long as the number of users K is smaller than the number N of transmit antennas, this procedure delivers excellent results. Our second approach is an iterative update algorithm for the max-min problem subject to a limitation of the transmit power. Given a low-complexity initialization beamformer, we search within the local vicinity of this filter vector for a filter-update preserving the transmit power and achieving a larger minimum SNR. To this end, we improve the weakest user's SNR during each iteration and keep on applying this procedure as long as the updates increase the smallest SNR. Otherwise, we adapt the step-size and continue investigating the local vicinity. It turns out that this novel approach is superior to existing state-of-the-art multicast beamformers for an arbitrary number of users.