Mesh extraction from occupancy functions is a useful tool in creating synthetic datasets for computer vision. However, existing mesh extraction methods have artifacts or performance profiles that limit their use. We propose OcMesher, a mesh extractor that efficiently handles high-detail unbounded scenes with perfect view consistency, with easy export to downstream real-time engines. The main novelty is an algorithm to construct an octree based on a given occupancy function and multiple camera views. We performed extensive experiments, and demonstrate OcMesher's usefulness for synthetic training & benchmark datasets, generating real-time environments for embodied AI and mesh extraction from depthmaps or novel view synthesis methods.

Building a foundation model for 3D vision is a complex challenge that remains unsolved. Towards that goal, it is important to understand the 3D reasoning capabilities of current models as well as identify the gaps between these models and humans. Therefore, we construct a new 3D visual understanding benchmark named UniQA-3D. UniQA-3D covers fundamental 3D vision tasks in the Visual Question Answering (VQA) format. We evaluate state-of-the-art Vision-Language Models (VLMs), specialized models, and human subjects on it. Our results show that VLMs generally perform poorly, while the specialized models are accurate but not robust, failing under geometric perturbations. In contrast, human vision continues to be the most reliable 3D visual system. We further demonstrate that neural networks align more closely with human 3D vision mechanisms compared to classical computer vision methods, and Transformer-based networks such as ViT [17] align more closely with human 3D vision mechanisms than CNNs. We hope our study will benefit the future development of foundation models for 3D vision. Code is available at https://github.com/princeton-vl/UniQA-3D.

We introduce Princeton365, a large-scale diverse dataset of 365 videos with accurate camera pose. Our dataset bridges the gap between accuracy and data diversity in current SLAM benchmarks by introducing a novel ground truth collection framework that leverages calibration boards and a 360-camera. We collect indoor, outdoor, and object scanning videos with synchronized monocular and stereo RGB video outputs as well as IMU. We further propose a new scene scale-aware evaluation metric for SLAM based on the optical flow induced by the camera pose estimation error. In contrast to the current metrics, our new metric allows for comparison between the performance of SLAM methods across scenes as opposed to existing metrics such as Average Trajectory Error (ATE), allowing researchers to analyze the failure modes of their methods. We also propose a challenging Novel View Synthesis benchmark that covers cases not covered by current NVS benchmarks, such as fully non-Lambertian scenes with 360-degree camera trajectories. Please visit https://princeton365.cs.princeton.edu for the dataset, code, videos, and submission.

Transparent objects are common in daily life, and understanding their multi-layer depth information -- perceiving both the transparent surface and the objects behind it -- is crucial for real-world applications that interact with transparent materials. In this paper, we introduce LayeredDepth, the first dataset with multi-layer depth annotations, including a real-world benchmark and a synthetic data generator, to support the task of multi-layer depth estimation. Our real-world benchmark consists of 1,500 images from diverse scenes, and evaluating state-of-the-art depth estimation methods on it reveals that they struggle with transparent objects. The synthetic data generator is fully procedural and capable of providing training data for this task with an unlimited variety of objects and scene compositions. Using this generator, we create a synthetic dataset with 15,300 images. Baseline models training solely on this synthetic dataset produce good cross-domain multi-layer depth estimation. Fine-tuning state-of-the-art single-layer depth models on it substantially improves their performance on transparent objects, with quadruplet accuracy on our benchmark increased from 55.14% to 75.20%. All images and validation annotations are available under CC0 at https://layereddepth.cs.princeton.edu.

Depth completion (DC) aims to predict a dense depth map from an RGB image and a sparse depth map. Existing DC methods generalize poorly to new datasets or unseen sparse depth patterns, limiting their real-world applications. We propose OMNI-DC, a highly robust DC model that generalizes well zero-shot to various datasets. The key design is a novel Multi-resolution Depth Integrator, allowing our model to deal with very sparse depth inputs. We also introduce a novel Laplacian loss to model the ambiguity in the training process. Moreover, we train OMNI-DC on a mixture of high-quality datasets with a scale normalization technique and synthetic depth patterns. Extensive experiments on 7 datasets show consistent improvements over baselines, reducing errors by as much as 43%. Codes and checkpoints are available at https://github.com/princeton-vl/OMNI-DC.

Vision Language Models (VLMs) excel at complex visual tasks such as VQA and chart understanding, yet recent work suggests they struggle with simple perceptual tests. We present an evaluation that tests vision-language models’ capacity for non-local visual reasoning— reasoning that requires chaining evidence collected from multiple, possibly distant, regions of an image. We isolate three distinct forms of non‑local vision: comparative perception, which demands holding two images in working memory and comparing them; saccadic search, which requires making discrete, evidence‑driven jumps to locate successive targets; and smooth visual search, which involves searching smoothly along a continuous contour. Flagship models (e.g., GPT-5, Gemini 2.5 Pro, Claude Sonnet 4), even those that perform well on prior primitive‑vision benchmarks, fail these tests and barely exceed random accuracy on two variants of our tasks that are trivial for humans. Our structured evaluation suite allows us to test if VLMs can perform similar visual algorithms to humans. Our findings show that despite gains in raw visual acuity, current models lack core visual reasoning capabilities.

Accurately tracking camera intrinsics is crucial for achieving 3D understanding from 2D video. However, most 3D algorithms assume that camera intrinsics stay constant throughout a video, which is often not true for many real-world in-the-wild videos. A major obstacle in this field is a lack of dynamic camera intrinsics benchmarks--existing benchmarks typically offer limited diversity in scene content and intrinsics variation, and none provide per-frame intrinsic changes for consecutive video frames. In this paper, we present Intrinsics in Flux (InFlux), a real-world benchmark that provides per-frame ground truth intrinsics annotations for videos with dynamic intrinsics. Compared to prior benchmarks, InFlux captures a wider range of intrinsic variations and scene diversity, featuring 143K+ annotated frames from 386 high-resolution indoor and outdoor videos with dynamic camera intrinsics. To ensure accurate per-frame intrinsics, we build a comprehensive lookup table of calibration experiments and extend the Kalibr toolbox to improve its accuracy and robustness. Using our benchmark, we evaluate existing baseline methods for predicting camera intrinsics and find that most struggle to achieve accurate predictions on videos with dynamic intrinsics. For the dataset, code, videos, and submission, please visit https://influx.cs.princeton.edu/.

Recent years have witnessed substantial progress on monocular depth estimation, particularly as measured by the success of large models on standard benchmarks. However, performance on standard benchmarks does not offer a complete assessment, because most evaluate accuracy but not robustness. In this work, we introduce PDE (Procedural Depth Evaluation), a new benchmark which enables systematic evaluation of robustness to changes in 3D scene content. PDE uses procedural generation to create 3D scenes that test robustness to various controlled perturbations, including object, camera, material and lighting changes. Our analysis yields interesting findings on what perturbations are challenging for state-of-the-art depth models, which we hope will inform further research. Code and data are available at https://github.com/princeton-vl/proc-depth-eval.

The procedural occupancy function is a flexible and compact representation for creating 3D scenes. For rasterization and other tasks, it is often necessary to extract a mesh that represents the shape. Unbounded scenes with long-range camera trajectories, such as flying through a forest, pose a unique challenge for mesh extraction. A single static mesh representing all the geometric detail necessary for the full camera path can be prohibitively large. Therefore, independent meshes can be extracted for different camera views, but this approach may lead to popping artifacts during transitions. We propose a temporally coherent method for extracting meshes suitable for long-range camera trajectories in unbounded scenes represented by an occupancy function. The key idea is to perform 4D mesh extraction using a new spacetime tree structure called a binary-octree. Experiments show that, compared to existing baseline methods, our method offers superior visual consistency at a comparable cost. The code and the supplementary video for this paper are available at https://github.com/princeton-vl/BinocMesher.

We introduce a new system for Multi-Session SLAM, which tracks camera motion across multiple disjoint videos under a single global reference. Our approach couples the prediction of optical flow with solver layers to estimate camera pose. The backbone is trained end-to-end using a novel differentiable solver for wide-baseline two-view pose. The full system can connect disjoint sequences, perform visualodometry, and global optimization. Compared to existing approaches, our design is accurate and robust to catas-trophic failures. Code is available at h t tps://github. com/princeton-v1/MultiSlam_DiffPose

We introduce Infinigen Indoors, a Blender-based procedural generator of photorealistic indoor scenes. It builds upon the existing Infinigen system, which focuses on natural scenes, but expands its coverage to indoor scenes by introducing a diverse library of procedural indoor assets, including furniture, architecture elements, appliances, and other day-to-day objects. It also introduces a constraint-based arrangement system, which consists of a domain-specific language for expressing diverse constraints on scene composition, and a solver that generates scene compositions that maximally satisfy the constraints. We provide an export tool that allows the generated 3D objects and scenes to be directly used for training embodied agents in real-time simulators such as Omniverse and Unreal. Infinigen Indoors is open-sourced under the BSD license. Please visit infinigen.org for code and videos.

We propose a new approach, Synthetic Optimized Layout with Instance Detection (SOLID), to pretrain object detectors with synthetic images. Our "SOLID" approach consists of two main components: (1) generating synthetic images using a collection of unlabelled 3D models with optimized scene arrangement; (2) pretraining an object detector on "instance detection" task—given a query image depicting an object, detecting all instances of the exact same object in a target image. Our approach does not need any semantic labels for pretraining and allows the use of arbitrary, diverse 3D models. Experiments on COCO show that with optimized data generation and a proper pretraining task, synthetic data can be highly effective data for pretraining object detectors. In particular, pretraining on rendered images achieves performance competitive with pretraining on real images while using significantly less computing resources. Code is available at https://github.com/princeton-vl/SOLID.

We introduce Infinigen, a procedural generator of photorealistic 3D scenes of the natural world. Infinigen is entirely procedural: every asset, from shape to texture, is generated from scratch via randomized mathematical rules, using no external source and allowing infinite variation and composition. Infinigen offers broad coverage of objects and scenes in the natural world including plants, animals, terrains, and natural phenomena such as fire, cloud, rain, and snow. Infinigen can be used to generate unlimited, diverse training data for a wide range of computer vision tasks including object detection, semantic segmentation, optical flow, and 3D reconstruction. We expect Infinigen to be a useful resource for computer vision research and beyond. Please visit infinigen.org for videos, code and pre-generated data.

In computer vision, 2D convolution is arguably the most important operation performed by a ConvNet. Unsurprisingly, it has been the focus of intense software and hardware optimization and enjoys highly efficient implementations. In this work, we ask an intriguing question: can we make a ConvNet work without 2D convolutions? Surprisingly, we find that the answer is yes—we show that a ConvNet consisting entirely of 1D convolutions can do just as well as 2D on ImageNet classification. Specifically, we find that one key ingredient to a high-performing 1D ConvNet is oriented 1D kernels: 1D kernels that are oriented not just horizontally or vertically, but also at other angles. Our experiments show that oriented 1D convolutions can not only replace 2D convolutions but also augment existing architectures with large kernels, leading to improved accuracy with minimal FLOPs increase. A key contribution of this work is a highly-optimized custom CUDA implementation of oriented 1D kernels, specialized to the depthwise convolution setting. Our benchmarks demonstrate that our custom CUDA implementation almost perfectly realizes the theoretical advantage of 1D convolution: it is faster than a native horizontal convolution for any arbitrary angle. Code is available at https://github.com/princeton-vl/Oriented1D.

Establishing correspondence between images or scenes is a significant challenge in computer vision, especially given occlusions, viewpoint changes, and varying object appearances. In this paper, we present Siamese Masked Autoencoders (SiamMAE), a simple extension of Masked Autoencoders (MAE) for learning visual correspondence from videos. SiamMAE operates on pairs of randomly sampled video frames and asymmetrically masks them. These frames are processed independently by an encoder network, and a decoder composed of a sequence of cross-attention layers is tasked with predicting the missing patches in the future frame. By masking a large fraction (95%) of patches in the future frame while leaving the past frame unchanged, SiamMAE encourages the network to focus on object motion and learn object-centric representations. Despite its conceptual simplicity, features learned via SiamMAE outperform state-of-the-art self-supervised methods on video object segmentation, pose keypoint propagation, and semantic part propagation tasks. SiamMAE achieves competitive results without relying on data augmentation, handcrafted tracking-based pretext tasks, or other techniques to prevent representational collapse.

We propose Deep Patch Visual Odometry (DPVO), a new deep learning system for monocular Visual Odometry (VO). DPVO uses a novel recurrent network architecture designed for tracking image patches across time. Recent approaches to VO have significantly improved the state-of-the-art accuracy by using deep networks to predict dense flow between video frames. However, using dense flow incurs a large computational cost, making these previous methods impractical for many use cases. Despite this, it has been assumed that dense flow is important as it provides additional redundancy against incorrect matches. DPVO disproves this assumption, showing that it is possible to get the best accuracy and efficiency by exploiting the advantages of sparse patch-based matching over dense flow. DPVO introduces a novel recurrent update operator for patch based correspondence coupled with differentiable bundle adjustment. On Standard benchmarks, DPVO outperforms all prior work, including the learning-based state-of-the-art VO-system (DROID) using a third of the memory while running 3x faster on average. Code is available at https://github.com/princeton-vl/DPVO

Accurate object rearrangement from vision is a crucial problem for a wide variety of real-world robotics applications in unstructured environments. We propose IFOR, Iterative Flow Minimization for Robotic Object Rearrangement, an end-to-end method for the challenging problem of object rearrangement for unknown objects given an RGBD image of the original and final scenes. First, we learn an optical flow model based on RAFT to estimate the relative transformation of the objects purely from synthetic data. This flow is then used in an iterative minimization algorithm to achieve accurate positioning of previously unseen objects. Crucially, we show that our method applies to cluttered scenes, and in the real world, while training only on synthetic data. Videos are available at h t t ps: //imankgoyal.github.io/ifor.html.

We address the task of 6D multi-object pose: given a set of known 3D objects and an RGB or RGB-D input image, we detect and estimate the 6D pose of each object. We propose a new approach to 6D object pose estimation which consists of an end-to-end differentiable architecture that makes use of geometric knowledge. Our approach iteratively refines both pose and correspondence in a tightly coupled manner, allowing us to dynamically remove outliers to improve accuracy. We use a novel differentiable layer to perform pose refinement by solving an optimization problem we refer to as Bidirectional Depth-Augmented Perspective-N-Point (BD-PnP). Our method achieves state-of-the-art accuracy on standard 6D Object Pose benchmarks. Code is available at https://github.com/princeton-vl/Coupled-Iterative-Refinement.

Reasoning over natural language is a challenging problem in NLP.In this work, we focus on proof generation: Given a hypothesis and a set of supporting facts, the model generates a proof tree indicating how to derive the hypothesis from supporting facts.Compared to generating the entire proof in one shot, stepwise generation can better exploit the compositionality and generalize to longer proofs but has achieved limited success on real-world data.Existing stepwise methods struggle to generate proof steps that are both logically valid and relevant to the hypothesis.Instead, they tend to hallucinate invalid steps given the hypothesis.In this paper, we present a novel stepwise method, NLProofS (Natural Language Proof Search), which learns to generate relevant steps conditioning on the hypothesis.At the core of our approach, we train an independent verifier to check the validity of the proof steps to prevent hallucination.Instead of generating steps greedily, we search for proofs maximizing a global proof score judged by the verifier.NL-ProofS achieves state-of-the-art performance on EntailmentBank and RuleTaker.Specifically, it improves the correctness of predicted proofs from 27.7% to 33.3% in the distractor setting of EntailmentBank, demonstrating the effectiveness of NLProofS in generating challenging human-authored proofs. 1

Face obfuscation (blurring, mosaicing, etc.) has been shown to be effective for privacy protection; nevertheless, object recognition research typically assumes access to complete, unobfuscated images. In this paper, we explore the effects of face obfuscation on the popular ImageNet challenge visual recognition benchmark. Most categories in the ImageNet challenge are not people categories; however, many incidental people appear in the images, and their privacy is a concern. We first annotate faces in the dataset. Then we demonstrate that face obfuscation has minimal impact on the accuracy of recognition models. Concretely, we benchmark multiple deep neural networks on obfuscated images and observe that the overall recognition accuracy drops only slightly (<= 1.0%). Further, we experiment with transfer learning to 4 downstream tasks (object recognition, scene recognition, face attribute classification, and object detection) and show that features learned on obfuscated images are equally transferable. Our work demonstrates the feasibility of privacy-aware visual recognition, improves the highly-used ImageNet challenge benchmark, and suggests an important path for future visual datasets. Data and code are available at https://github.com/princetonvisualai/imagenet-face-obfuscation.

Latency is of utmost importance in safety-critical systems. In neural networks, lowest theoretical latency is dependent on the depth of the network. This begs the question -- is it possible to build high-performing ``non-deep" neural networks? We show that it is. To do so, we use parallel subnetworks instead of stacking one layer after another. This helps effectively reduce depth while maintaining high performance. By utilizing parallel substructures, we show, for the first time, that a network with a depth of just 12 can achieve top-1 accuracy over 80% on ImageNet, 96% on CIFAR10, and 81% on CIFAR100. We also show that a network with a low-depth (12) backbone can achieve an AP of 48% on MS-COCO. We analyze the scaling rules for our design and show how to increase performance without changing the network's depth. Finally, we provide a proof of concept for how non-deep networks could be used to build low-latency recognition systems. Code is available at https://github.com/imankgoyal/NonDeepNetworks.

We introduce RAFT-Stereo, a new deep architecture for rectified stereo based on the optical flow network RAFT [35]. We introduce multi-level convolutional GRUs, which more efficiently propagate information across the image. A modified version of RAFT-Stereo can perform accurate real-time inference. RAFT-stereo ranks first on the Middlebury leaderboard, outperforming the next best method on 1px error by 29% and outperforms all published work on the ETH3D two-view stereo benchmark. Code is available at https://github.com/princeton-vl/RAFT-Stereo.

Recent progress on physics-based character animation has shown impressive breakthroughs on human motion synthesis, through imitating motion capture data via deep reinforcement learning. However, results have mostly been demonstrated on imitating a single distinct motion pattern, and do not generalize to interactive tasks that require flexible motion patterns due to varying human-object spatial configurations. To bridge this gap, we focus on one class of interactive tasks---sitting onto a chair. We propose a hierarchical reinforcement learning framework which relies on a collection of subtask controllers trained to imitate simple, reusable mocap motions, and a meta controller trained to execute the subtasks properly to complete the main task. We experimentally demonstrate the strength of our approach over different non-hierarchical and hierarchical baselines. We also show that our approach can be applied to motion prediction given an image input. A supplementary video can be found at https://youtu.be/3CeN0OGz2cA.

Recent work introduced progressive network growing as a promising way to ease the training for large GANs, but the model design and architecture-growing strategy still remain under-explored and needs manual design for different image data. In this paper, we propose a method to dynamically grow a GAN during training, optimizing the network architecture and its parameters together with automation. The method embeds architecture search techniques as an interleaving step with gradient-based training to periodically seek the optimal architecture-growing strategy for the generator and discriminator. It enjoys the benefits of both eased training because of progressive growing and improved performance because of broader architecture design space. Experimental results demonstrate new state-of-the-art of image generation. Observations in the search procedure also provide constructive insights into the GAN model design such as generator-discriminator balance and convolutional layer choices.

We address the problem of scene flow: given a pair of stereo or RGB-D video frames, estimate pixelwise 3D motion. We introduce RAFT-3D, a new deep architecture for scene flow. RAFT-3D is based on the RAFT model developed for optical flow but iteratively updates a dense field of pixelwise SE3 motion instead of 2D motion. A key innovation of RAFT-3D is rigid-motion embeddings, which represent a soft grouping of pixels into rigid objects. Integral to rigid-motion embeddings is Dense-SE3, a differentiable layer that enforces geometric consistency of the embeddings. Experiments show that RAFT-3D achieves state-of-the-art performance. On FlyingThings3D, under the twoview evaluation, we improved the best published accuracy (δ < 0.05) from 34.3% to 83.7%. On KITTI, we achieve an error of 5.77, outperforming the best published method (6.31), despite using no object instance supervision.

We address the problem of performing backpropagation for computation graphs involving 3D transformation groups SO(3), SE(3), and Sim(3). 3D transformation groups are widely used in 3D vision and robotics, but they do not form vector spaces and instead lie on smooth manifolds. The standard backpropagation approach, which embeds 3D transformations in Euclidean spaces, suffers from numerical difficulties. We introduce a new library, which exploits the group structure of 3D transformations and performs backpropagation in the tangent spaces of manifolds. We show that our approach is numerically more stable, easier to implement, and beneficial to a diverse set of tasks. Our plug-and-play PyTorch library is available at https://github.com/princeton-vl/lietorch.

Processing point cloud data is an important component of many real-world systems. As such, a wide variety of point-based approaches have been proposed, reporting steady benchmark improvements over time. We study the key ingredients of this progress and uncover two critical results. First, we find that auxiliary factors like different evaluation schemes, data augmentation strategies, and loss functions, which are independent of the model architecture, make a large difference in performance. The differences are large enough that they obscure the effect of architecture. When these factors are controlled for, PointNet++, a relatively older network, performs competitively with recent methods. Second, a very simple projection-based method, which we refer to as SimpleView, performs surprisingly well. It achieves on par or better results than sophisticated state-of-the-art methods on ModelNet40 while being half the size of PointNet++. It also outperforms state-of-the-art methods on ScanObjectNN, a real-world point cloud benchmark, and demonstrates better cross-dataset generalization. Code is available at https://github.com/princeton-vl/SimpleView.

We introduce Recurrent All-Pairs Field Transforms (RAFT), a new deep network architecture for optical flow. RAFT extracts per-pixel features, builds multi-scale 4D correlation volumes for all pairs of pixels, and iteratively updates a flow field through a recurrent unit that performs lookups on the correlation volumes. RAFT achieves state-of-the-art performance on the KITTI and Sintel datasets. In addition, RAFT has strong cross-dataset generalization as well as high efficiency in inference time, training speed, and parameter count.

We introduce DROID-SLAM, a new deep learning based SLAM system. DROID-SLAM consists of recurrent iterative updates of camera pose and pixelwise depth through a Dense Bundle Adjustment layer. DROID-SLAM is accurate, achieving large improvements over prior work, and robust, suffering from substantially fewer catastrophic failures. Despite training on monocular video, it can leverage stereo or RGB-D video to achieve improved performance at test time. The URL to our open source code is https://github.com/princeton-vl/DROID-SLAM.

Single-view 3D is the task of recovering 3D properties such as depth and surface normals from a single image. We hypothesize that a major obstacle to single-image 3D is data. We address this issue by presenting Open Annotations of Single Image Surfaces (OASIS), a dataset for single-image 3D in the wild consisting of annotations of detailed 3D geometry for 140,000 images. We train and evaluate leading models on a variety of single-image 3D tasks. We expect OASIS to be a useful resource for 3D vision research. Project site: https://pvl.cs.princeton.edu/OASIS.

Recent advances have spurred incredible progress in self-supervised pretraining for vision. We investigate what factors may play a role in the utility of these pretraining methods for practitioners. To do this, we evaluate various self-supervised algorithms across a comprehensive array of synthetic datasets and downstream tasks. We prepare a suite of synthetic data that enables an endless supply of annotated images as well as full control over dataset difficulty. Our experiments offer insights into how the utility of self-supervision changes as the number of available labels grows as well as how the utility changes as a function of the downstream task and the properties of the training data. We also find that linear evaluation does not correlate with finetuning performance. Code and data is available at github.com/princeton-vl/selfstudy.

Synthetic images rendered by graphics engines are a promising source for training deep networks. However, it is challenging to ensure that they can help train a network to perform well on real images, because a graphics-based generation pipeline requires numerous design decisions such as the selection of 3D shapes and the placement of the camera. In this work, we propose a new method that optimizes the generation of 3D training data based on what we call "hybrid gradient". We parametrize the design decisions as a real vector, and combine the approximate gradient and the analytical gradient to obtain the hybrid gradient of the network performance with respect to this vector. We evaluate our approach on the task of estimating surface normal, depth or intrinsic decomposition from a single image. Experiments on standard benchmarks show that our approach can outperform the prior state of the art on optimizing the generation of 3D training data, particularly in terms of computational efficiency.

The ability to jointly understand the geometry of objects and plan actions for manipulating them is crucial for intelligent agents. We refer to this ability as geometric planning. Recently, many interactive environments have been proposed to evaluate intelligent agents on various skills, however, none of them cater to the needs of geometric planning. We present PackIt, a virtual environment to evaluate and potentially learn the ability to do geometric planning, where an agent needs to take a sequence of actions to pack a set of objects into a box with limited space. We also construct a set of challenging packing tasks using an evolutionary algorithm. Further, we study various baselines for the task that include model-free learning-based and heuristic-based methods, as well as search-based optimization methods that assume access to the model of the environment.

We introduce LifeQA, a benchmark dataset for video question answering that focuses on day-to-day real-life situations. Current video question answering datasets consist of movies and TV shows. However, it is well-known that these visual domains are not representative of our day-to-day lives. Movies and TV shows, for example, benefit from professional camera movements, clean editing, crisp audio recordings, and scripted dialog between professional actors. While these domains provide a large amount of data for training models, their properties make them unsuitable for testing real-life question answering systems. Our dataset, by contrast, consists of video clips that represent only real-life scenarios. We collect 275 such video clips and over 2.3k multiple-choice questions. In this paper, we analyze the challenging but realistic aspects of LifeQA, and we apply several state-of-the-art video question answering models to provide benchmarks for future research. The full dataset is publicly available at https://lit.eecs.umich.edu/lifeqa/.

Understanding spatial relations (e.g., laptop on table) in visual input is important for both humans and robots. Existing datasets are insufficient as they lack large-scale, high-quality 3D ground truth information, which is critical for learning spatial relations. In this paper, we fill this gap by constructing Rel3D: the first large-scale, human-annotated dataset for grounding spatial relations in 3D. Rel3D enables quantifying the effectiveness of 3D information in predicting spatial relations on large-scale human data. Moreover, we propose minimally contrastive data collection---a novel crowdsourcing method for reducing dataset bias. The 3D scenes in our dataset come in minimally contrastive pairs: two scenes in a pair are almost identical, but a spatial relation holds in one and fails in the other. We empirically validate that minimally contrastive examples can diagnose issues with current relation detection models as well as lead to sample-efficient training. Code and data are available at https://github.com/princeton-vl/Rel3D.

We consider the task of automated theorem proving, a key AI task. Deep learning has shown promise for training theorem provers, but there are limited human-written theorems and proofs available for supervised learning. To address this limitation, we propose to learn a neural generator that automatically synthesizes theorems and proofs for the purpose of training a theorem prover. Experiments on real-world tasks demonstrate that synthetic data from our approach improves the theorem prover and advances the state of the art of automated theorem proving in Metamath.

Parsing sentences into syntax trees can benefit downstream applications in NLP. Transition-based parsers build trees by executing actions in a state transition system. They are computationally efficient, and can leverage machine learning to predict actions based on partial trees. However, existing transition-based parsers are predominantly based on the shift-reduce transition system, which does not align with how humans are known to parse sentences. Psycholinguistic research suggests that human parsing is strongly incremental—humans grow a single parse tree by adding exactly one token at each step. In this paper, we propose a novel transition system called attach-juxtapose. It is strongly incremental; it represents a partial sentence using a single tree; each action adds exactly one token into the partial tree. Based on our transition system, we develop a strongly incremental parser. At each step, it encodes the partial tree using a graph neural network and predicts an action. We evaluate our parser on Penn Treebank (PTB) and Chinese Treebank (CTB). On PTB, it outperforms existing parsers trained with only constituency trees; and it performs on par with state-of-the-art parsers that use dependency trees as additional training data. On CTB, our parser establishes a new state of the art. Code is available at https://github.com/princeton-vl/attach-juxtapose-parser.

State-of-the-art methods for action recognition commonly use two networks: the spatial stream, which takes RGB frames as input, and the temporal stream, which takes optical flow as input. In recent work, both streams are 3D Convolutional Neural Networks, which use spatiotemporal filters. These filters can respond to motion, and therefore should allow the network to learn motion representations, removing the need for optical flow. However, we still see significant benefits in performance by feeding optical flow into the temporal stream, indicating that the spatial stream is "missing" some of the signal that the temporal stream captures. In this work, we first investigate whether motion representations are indeed missing in the spatial stream, and show that there is significant room for improvement. Second, we demonstrate that these motion representations can be improved using distillation, that is, by tuning the spatial stream to mimic the temporal stream, effectively combining both models into a single stream. Finally, we show that our Distilled 3D Network (D3D) achieves performance on par with the two-stream approach, with no need to compute optical flow during inference.

We consider the task of identifying human actions visible in online videos. We focus on the widely spread genre of lifestyle vlogs, which consist of videos of people performing actions while verbally describing them. Our goal is to identify if actions mentioned in the speech description of a video are visually present. We construct a dataset with crowdsourced manual annotations of visible actions, and introduce a multimodal algorithm that leverages information derived from visual and linguistic clues to automatically infer which actions are visible in a video. We demonstrate that our multimodal algorithm outperforms algorithms based only on one modality at a time.

We introduce a new embedding model to represent movie characters and their interactions in a dialogue by encoding in the same representation the language used by these characters as well as information about the other participants in the dialogue.We evaluate the performance of these new character embeddings on two tasks: (1) character relatedness, using a dataset we introduce consisting of a dense character interaction matrix for 4,761 unique character pairs over 22 hours of dialogue from eighteen movies; and (2) character relation classification, for fine-and coarse-grained relations, as well as sentiment relations.Our experiments show that our model significantly outperforms the traditional Word2Vec continuous bag-of-words and skip-gram models, demonstrating the effectiveness of the character embeddings we introduce.We further show how these embeddings can be used in conjunction with a visual question answering system to improve over previous results.

Depth estimation from a single image in the wild remains a challenging problem. One main obstacle is the lack of high-quality training data for images in the wild. In this paper we propose a method to automatically generate such data through Structure-from-Motion (SfM) on Internet videos. The core of this method is a Quality Assessment Network that identifies high-quality reconstructions obtained from SfM. Using this method, we collect single-view depth training data from a large number of YouTube videos and construct a new dataset called YouTube3D. Experiments show that YouTube3D is useful in training depth estimation networks and advances the state of the art of single-view depth estimation in the wild.

Highly expressive models such as deep neural networks (DNNs) have been widely applied to various applications. However, recent studies show that DNNs are vulnerable to adversarial examples, which are carefully crafted inputs aiming to mislead the predictions. Currently, the majority of these studies have focused on perturbation added to image pixels, while such manipulation is not physically realistic. Some works have tried to overcome this limitation by attaching printable 2D patches or painting patterns onto surfaces, but can be potentially defended because 3D shape features are intact. In this paper, we propose meshAdv to generate "adversarial 3D meshes" from objects that have rich shape features but minimal textural variation. To manipulate the shape or texture of the objects, we make use of a differentiable renderer to compute accurate shading on the shape and propagate the gradient. Extensive experiments show that the generated 3D meshes are effective in attacking both classifiers and object detectors. We evaluate the attack under different viewpoints. In addition, we design a pipeline to perform black-box attack on a photorealistic renderer with unknown rendering parameters.

This paper explores object detection in the small data regime, where only a limited number of annotated bounding boxes are available due to data rarity and annotation expense. This is a common challenge today with machine learning being applied to many new tasks where obtaining training data is more challenging, e.g. in medical images with rare diseases that doctors sometimes only see once in their life-time. In this work we explore this problem from a generative modeling perspective by learning to generate new images with associated bounding boxes, and using these for training an object detector. We show that simply training previously proposed generative models does not yield satisfactory performance due to them optimizing for image realism rather than object detection accuracy. To this end we develop a new model with a novel unrolling mechanism that jointly optimizes the generative model and a detector such that the generated images improve the performance of the detector. We show this method outperforms the state of the art on two challenging datasets, disease detection and small data pedestrian detection, improving the average precision on NIH Chest X-ray by a relative 20% and localization accuracy by a relative 50%.

Understanding the spatial relations between objects in images is a surprisingly challenging task. A chair may be "behind" a person even if it appears to the left of the person in the image (depending on which way the person is facing). Two students that appear close to each other in the image may not in fact be "next to" each other if there is a third student between them. We introduce SpatialSense, a dataset specializing in spatial relation recognition which captures a broad spectrum of such challenges, allowing for proper benchmarking of computer vision techniques. SpatialSense is constructed through adversarial crowdsourcing, in which human annotators are tasked with finding spatial relations that are difficult to predict using simple cues such as 2D spatial configuration or language priors. Adversarial crowdsourcing significantly reduces dataset bias and samples more interesting relations in the long tail compared to existing datasets. On SpatialSense, state-of-the-art recognition models perform comparably to simple baselines, suggesting that they rely on straightforward cues instead of fully reasoning about this complex task. The SpatialSense benchmark provides a path forward to advancing the spatial reasoning capabilities of computer vision systems. The dataset and code are available at https://github.com/princeton-vl/SpatialSense.

Humans prove theorems by relying on substantial high-level reasoning and problem-specific insights. Proof assistants offer a formalism that resembles human mathematical reasoning, representing theorems in higher-order logic and proofs as high-level tactics. However, human experts have to construct proofs manually by entering tactics into the proof assistant. In this paper, we study the problem of using machine learning to automate the interaction with proof assistants. We construct CoqGym, a large-scale dataset and learning environment containing 71K human-written proofs from 123 projects developed with the Coq proof assistant. We develop ASTactic, a deep learning-based model that generates tactics as programs in the form of abstract syntax trees (ASTs). Experiments show that ASTactic trained on CoqGym can generate effective tactics and can be used to prove new theorems not previously provable by automated methods. Code is available at https://github.com/princeton-vl/CoqGym.

We introduce Dynamic Deep Neural Networks (D2NN), a new type of feed-forward deep neural network that allows selective execution. Given an input, only a subset of D2NN neurons are executed, and the particular subset is determined by the D2NN itself. By pruning unnecessary computation depending on input, D2NNs provide a way to improve computational efficiency. To achieve dynamic selective execution, a D2NN augments a feed-forward deep neural network (directed acyclic graph of differentiable modules) with controller modules. Each controller module is a sub-network whose output is a decision that controls whether other modules can execute. A D2NN is trained end to end. Both regular and controller modules in a D2NN are learnable and are jointly trained to optimize both accuracy and efficiency. Such training is achieved by integrating backpropagation with reinforcement learning. With extensive experiments of various D2NN architectures on image classification tasks, we demonstrate that D2NNs are general and flexible, and can effectively optimize accuracy-efficiency trade-offs.

In this paper, we study the problem of geometric reasoning in the context of question-answering.We introduce Dynamic Spatial Memory Network (DSMN), a new deep network architecture designed for answering questions that admit latent visual representations.DSMN learns to generate and reason over such representations.Further, we propose two synthetic benchmarks, FloorPlanQA and ShapeIntersection, to evaluate the geometric reasoning capability of QA systems.Experimental results validate the effectiveness of our proposed DSMN for visual thinking tasks 1 .

We propose TAL-Net, an improved approach to temporal action localization in video that is inspired by the Faster RCNN object detection framework. TAL-Net addresses three key shortcomings of existing approaches: (1) we improve receptive field alignment using a multi-scale architecture that can accommodate extreme variation in action durations; (2) we better exploit the temporal context of actions for both proposal generation and action classification by appropriately extending receptive fields; and (3) we explicitly consider multi-stream feature fusion and demonstrate that fusing motion late is important. We achieve state-of-the-art performance for both action proposal and localization on THUMOS'14 detection benchmark and competitive performance on ActivityNet challenge.

Batch Normalization (BN) is capable of accelerating the training of deep models by centering and scaling activations within mini-batches. In this work, we propose Decorrelated Batch Normalization (DBN), which not just centers and scales activations but whitens them. We explore multiple whitening techniques, and find that PCA whitening causes a problem we call stochastic axis swapping, which is detrimental to learning. We show that ZCA whitening does not suffer from this problem, permitting successful learning. DBN retains the desirable qualities of BN and further improves BN's optimization efficiency and generalization ability. We design comprehensive experiments to show that DBN can improve the performance of BN on multilayer perceptrons and convolutional neural networks. Furthermore, we consistently improve the accuracy of residual networks on CIFAR-10, CIFAR-100, and ImageNet.

In this paper, we address the shape-from-shading problem by training deep networks with synthetic images. Unlike conventional approaches that combine deep learning and synthetic imagery, we propose an approach that does not need any external shape dataset to render synthetic images. Our approach consists of two synergistic processes: the evolution of complex shapes from simple primitives, and the training of a deep network for shape-from-shading. The evolution generates better shapes guided by the network training, while the training improves by using the evolved shapes. We show that our approach achieves state-of-the-art performance on a shape-from-shading benchmark.

Mahmoud Azab, Mingzhe Wang, Max Smith, Noriyuki Kojima, Jia Deng, Rada Mihalcea. Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers). 2018.

We study the problem of detecting human-object interactions (HOI) in static images, defined as predicting a human and an object bounding box with an interaction class label that connects them. HOI detection is a fundamental problem in computer vision as it provides semantic information about the interactions among the detected objects. We introduce HICO-DET, a new large benchmark for HOI detection, by augmenting the current HICO classification benchmark with instance annotations. To solve the task, we propose Human-Object Region-based Convolutional Neural Networks (HO-RCNN). At the core of our HO-RCNN is the Interaction Pattern, a novel DNN input that characterizes the spatial relations between two bounding boxes. Experiments on HICO-DET demonstrate that our HO-RCNN, by exploiting human-object spatial relations through Interaction Patterns, significantly improves the performance of HOI detection over baseline approaches.

Targeted socio-economic policies require an accurate understanding of a country’s demographic makeup. To that end, the United States spends more than 1 billion dollars a year gathering census data such as race, gender, education, occupation and unemployment rates. Compared to the traditional method of collecting surveys across many years which is costly and labor intensive, data-driven, machine learning-driven approaches are cheaper and faster—with the potential ability to detect trends in close to real time. In this work, we leverage the ubiquity of Google Street View images and develop a computer vision pipeline to predict income, per capita carbon emission, crime rates and other city attributes from a single source of publicly available visual data. We first detect cars in 50 million images across 200 of the largest US cities and train a model to predict demographic attributes using the detected cars. To facilitate our work, we have collected the largest and most challenging fine-grained dataset reported to date consisting of over 2600 classes of cars comprised of images from Google Street View and other web sources, classified by car experts to account for even the most subtle of visual differences. We use this data to construct the largest scale fine-grained detection system reported to date. Our prediction results correlate well with ground truth income data (r=0.82), Massachusetts department of vehicle registration, and sources investigating crime rates, income segregation, per capita carbon emission, and other market research. Finally, we learn interesting relationships between cars and neighborhoods allowing us to perform the first large scale sociological analysis of cities using computer vision techniques.

We present a crowdsourcing workflow to collect image annotations for visually similar synthetic categories without requiring experts. In animals, there is a direct link between taxonomy and visual similarity: e.g. a collie (type of dog) looks more similar to other collies (e.g. smooth collie) than a greyhound (another type of dog). However, in synthetic categories such as cars, objects with similar taxonomy can have very different appearance: e.g. a 2011 Ford F-150 Supercrew-HD looks the same as a 2011 Ford F-150 Supercrew-LL but very different from a 2011 Ford F-150 Supercrew-SVT. We introduce a graph based crowdsourcing algorithm to automatically group visually indistinguishable objects together. Using our workflow, we label 712,430 images by ~1,000 Amazon Mechanical Turk workers; resulting in the largest fine-grained visual dataset reported to date with 2,657 categories of cars annotated at 1/20th the cost of hiring experts.

This paper presents the first study on forecasting human dynamics from static images. The problem is to input a single RGB image and generate a sequence of upcoming human body poses in 3D. To address the problem, we propose the 3D Pose Forecasting Network (3D-PFNet). Our 3D-PFNet integrates recent advances on single-image human pose estimation and sequence prediction, and converts the 2D predictions into 3D space. We train our 3D-PFNet using a three-step training strategy to leverage a diverse source of training data, including image and video based human pose datasets and 3D motion capture (MoCap) data. We demonstrate competitive performance of our 3D-PFNet on 2D pose forecasting and 3D structure recovery through quantitative and qualitative results.

We address the problem of temporal action localization in videos. We pose action localization as a structured prediction over arbitrary-length temporal windows, where each window is scored as the sum of frame-wise classification scores. Additionally, our model classifies the start, middle, and end of each action as separate components, allowing our system to explicitly model each actions temporal evolution and take advantage of informative temporal dependencies present in that structure. In this framework, we localize actions by searching for the structured maximal sum, a problem for which we develop a novel, provably-efficient algorithmic solution. The frame-wise classification scores are computed using features from a deep Convolutional Neural Network (CNN), which are trained end-to-end to directly optimize for a novel structured objective. We evaluate our system on the THUMOS 14 action detection benchmark and achieve competitive performance.

We study the problem of single-image depth estimation for images in the wild. We collect human annotated surface normals and use them to help train a neural network that directly predicts pixel-wise depth. We propose two novel loss functions for training with surface normal annotations. Experiments on NYU Depth, KITTI, and our own dataset demonstrate that our approach can significantly improve the quality of depth estimation in the wild.

Graphs are a useful abstraction of image content. Not only can graphs represent details about individual objects in a scene but they can capture the interactions between pairs of objects. We present a method for training a convolutional neural network such that it takes in an input image and produces a full graph definition. This is done end-to-end in a single stage with the use of associative embeddings. The network learns to simultaneously identify all of the elements that make up a graph and piece them together. We benchmark on the Visual Genome dataset, and demonstrate state-of-the-art performance on the challenging task of scene graph generation.

We introduce associative embedding, a novel method for supervising convolutional neural networks for the task of detection and grouping. A number of computer vision problems can be framed in this manner including multi-person pose estimation, instance segmentation, and multi-object tracking. Usually the grouping of detections is achieved with multi-stage pipelines, instead we propose an approach that teaches a network to simultaneously output detections and group assignments. This technique can be easily integrated into any state-of-the-art network architecture that produces pixel-wise predictions. We show how to apply this method to multi-person pose estimation and report state-of-the-art performance on the MPII and MS-COCO datasets.

We propose a deep learning-based approach to the problem of premise selection: selecting mathematical statements relevant for proving a given conjecture. We represent a higher-order logic formula as a graph that is invariant to variable renaming but still fully preserves syntactic and semantic information. We then embed the graph into a vector via a novel embedding method that preserves the information of edge ordering. Our approach achieves state-of-the-art results on the HolStep dataset, improving the classification accuracy from 83% to 90.3%.

This paper studies single-image depth perception in the wild, i.e., recovering depth from a single image taken in unconstrained settings. We introduce a new dataset “Depth in the Wild” consisting of images in the wild annotated with relative depth between pairs of random points. We also propose a new algorithm that learns to estimate metric depth using annotations of relative depth. Compared to the state of the art, our algorithm is simpler and performs better. Experiments show that our algorithm, combined with existing RGB-D data and our new relative depth annotations, significantly improves single-image depth perception in the wild.

Fine-grained recognition concerns categorization at sub-ordinate levels, where the distinction between object classes is highly local. Compared to basic level recognition, fine-grained categorization can be more challenging as there are in general less data and fewer discriminative features. This necessitates the use of a stronger prior for feature selection. In this work, we include humans in the loop to help computers select discriminative features. We introduce a novel online game called "Bubbles" that reveals discriminative features humans use. The player's goal is to identify the category of a heavily blurred image. During the game, the player can choose to reveal full details of circular regions ("bubbles"), with a certain penalty. With proper setup the game generates discriminative bubbles with assured quality. We next propose the "BubbleBank" representation that uses the human selected bubbles to improve machine recognition performance. Finally, we demonstrate how to extend BubbleBank to a view-invariant 3D representation. Experiments demonstrate that our approach yields large improvements over the previous state of the art on challenging benchmarks.

Affordances are fundamental attributes of objects. Affordances reveal the functionalities of objects and the possible actions that can be performed on them. Understanding affordances is crucial for recognizing human activities in visual data and for robots to interact with the world. In this paper we introduce the new problem of mining the knowledge of semantic affordance: given an object, determining whether an action can be performed on it. This is equivalent to connecting verb nodes and noun nodes in WordNet, or filling an affordance matrix encoding the plausibility of each action-object pair. We introduce a new benchmark with crowdsourced ground truth affordances on 20 PASCAL VOC object classes and 957 action classes. We explore a number of approaches including text mining, visual mining, and collaborative filtering. Our analyses yield a number of significant insights that reveal the most effective ways of collecting knowledge of semantic affordances.

Human actions capture a wide variety of interactions between people and objects. As a result, the set of possible actions is extremely large and it is difficult to obtain sufficient training examples for all actions. However, we could compensate for this sparsity in supervision by leveraging the rich semantic relationship between different actions. A single action is often composed of other smaller actions and is exclusive of certain others. We need a method which can reason about such relationships and extrapolate unobserved actions from known actions. Hence, we propose a novel neural network framework which jointly extracts the relationship between actions and uses them for training better action retrieval models. Our model incorporates linguistic, visual and logical consistency based cues to effectively identify these relationships. We train and test our model on a largescale image dataset of human actions. We show a significant improvement in mean AP compared to different baseline methods including the HEX-graph approach from Deng et al. [8].

We introduce a new benchmark "Humans Interacting with Common Objects" (HICO) for recognizing human-object interactions (HOI). We demonstrate the key features of HICO: a diverse set of interactions with common object categories, a list of well-defined, sense-based HOI categories, and an exhaustive labeling of co-occurring interactions with an object category in each image. We perform an in-depth analysis of representative current approaches and show that DNNs enjoy a significant edge. In addition, we show that semantic knowledge can significantly improve HOI recognition, especially for uncommon categories.

We consider classification problems in which the label space has structure. A common example is hierarchical label spaces, corresponding to the case where one label subsumes another (e.g., animal subsumes dog). But labels can also be mutually exclusive (e.g., dog vs cat) or unrelated (e.g., furry, carnivore). To jointly model hierarchy and exclusion relations, the notion of a HEX (hierarchy and exclusion) graph was introduced in [8]. This combined a conditional random field (CRF) with a deep neural network (DNN), resulting in state of the art results when applied to visual object classification problems where the training labels were drawn from different levels of the ImageNet hierarchy (e.g., an image might be labeled with the basic level category "dog", rather than the more specific label "husky"). In this paper, we extend the HEX model to allow for soft or probabilistic relations between labels, which is useful when there is uncertainty about the relationship between two labels (e.g., an antelope is "sort of" furry, but not to the same degree as a grizzly bear). We call our new model pHEX, for probabilistic HEX. We show that the pHEX graph can be converted to an Ising model, which allows us to use existing off-the-shelf inference methods (in contrast to the HEX method, which needed specialized inference algorithms). Experimental results show significant improvements in a number of large-scale visual object classification tasks, outperforming the previous HEX model.

We study strategies for scalable multi-label annotation, or for efficiently acquiring multiple labels from humans for a collection of items. We propose an algorithm that exploits correlation, hierarchy, and sparsity of the label distribution. A case study of labeling 200 objects using 20,000 images demonstrates the effectiveness of our approach. The algorithm results in up to 6x reduction in human computation time compared to the naive method of querying a human annotator for the presence of every object in every image.

This paper addresses the problem of fine-grained recognition: recognizing subordinate categories such as bird species, car models, or dog breeds. We focus on two major challenges: learning expressive appearance descriptors and localizing discriminative parts. To this end, we propose an object representation that detects important parts and describes fine grained appearances. The part detectors are learned in a fully unsupervised manner, based on the insight that images with similar poses can be automatically discovered for fine-grained classes in the same domain. The appearance descriptors are learned using a convolutional neural network. Our approach requires only image level class labels, without any use of part annotations or segmentation masks, which may be costly to obtain. We show experimentally that combining these two insights is an effective strategy for fine-grained recognition.

Fine-grained recognition concerns categorization at sub-ordinate levels, where the distinction between object classes is highly local. Compared to basic level recognition, fine-grained categorization can be more challenging as there are in general less data and fewer discriminative features. This necessitates the use of stronger prior for feature selection. In this work, we include humans in the loop to help computers select discriminative features. We introduce a novel online game called "Bubbles" that reveals discriminative features humans use. The player's goal is to identify the category of a heavily blurred image. During the game, the player can choose to reveal full details of circular regions ("bubbles"), with a certain penalty. With proper setup the game generates discriminative bubbles with assured quality. We next propose the "Bubble Bank" algorithm that uses the human selected bubbles to improve machine recognition performance. Experiments demonstrate that our approach yields large improvements over the previous state of the art on challenging benchmarks.

Entry level categories - the labels people will use to name an object - were originally defined and studied by psychologists in the 1980s. In this paper we study entry-level categories at a large scale and learn the first models for predicting entry-level categories for images. Our models combine visual recognition predictions with proxies for word "naturalness" mined from the enormous amounts of text on the web. We demonstrate the usefulness of our models for predicting nouns (entry-level words) associated with images by people. We also learn mappings between concepts predicted by existing visual recognition systems and entry-level concepts that could be useful for improving human-focused applications such as natural language image description or retrieval.

The growth of detection datasets and the multiple directions of object detection research provide both an unprecedented need and a great opportunity for a thorough evaluation of the current state of the field of categorical object detection. In this paper we strive to answer two key questions. First, where are we currently as a field: what have we done right, what still needs to be improved? Second, where should we be going in designing the next generation of object detectors? Inspired by the recent work of Hoiem et al. on the standard PASCAL VOC detection dataset, we perform a large-scale study on the Image Net Large Scale Visual Recognition Challenge (ILSVRC) data. First, we quantitatively demonstrate that this dataset provides many of the same detection challenges as the PASCAL VOC. Due to its scale of 1000 object categories, ILSVRC also provides an excellent test bed for understanding the performance of detectors as a function of several key properties of the object classes. We conduct a series of analyses looking at how different detection methods perform on a number of image-level and object-class-level properties such as texture, color, deformation, and clutter. We learn important lessons of the current object detection methods and propose a number of insights for designing the next generation object detectors.

As visual recognition scales up to ever larger numbers of categories, maintaining high accuracy is increasingly difficult. In this work, we study the problem of optimizing accuracy-specificity trade-offs in large scale recognition, motivated by the observation that object categories form a semantic hierarchy consisting of many levels of abstraction. A classifier can select the appropriate level, trading off specificity for accuracy in case of uncertainty. By optimizing this trade-off, we obtain classifiers that try to be as specific as possible while guaranteeing an arbitrarily high accuracy. We formulate the problem as maximizing information gain while ensuring a fixed, arbitrarily small error rate with a semantic hierarchy. We propose the Dual Accuracy Reward Trade-off Search (DARTS) algorithm and prove that, under practical conditions, it converges to an optimal solution. Experiments demonstrate the effectiveness of our algorithm on datasets ranging from 65 to over 10,000 categories.

This paper addresses the problem of similar image retrieval, especially in the setting of large-scale datasets with millions to billions of images. The core novel contribution is an approach that can exploit prior knowledge of a semantic hierarchy. When semantic labels and a hierarchy relating them are available during training, significant improvements over the state of the art in similar image retrieval are attained. While some of this advantage comes from the ability to use additional information, experiments exploring a special case where no additional data is provided, show the new approach can still outperform OASIS, the current state of the art for similarity learning. Exploiting hierarchical relationships is most important for larger scale problems, where scalability becomes crucial. The proposed learning approach is fundamentally parallelizable and as a result scales more easily than previous work. An additional contribution is a novel hashing scheme (for bilinear similarity on vectors of probabilities, optionally taking into account hierarchy) that is able to reduce the computational cost of retrieval. Experiments are performed on Caltech256 and the larger ImageNet dataset.

We present a novel approach to efficiently learn a label tree for large scale classification with many classes. The key contribution of the approach is a technique to simultaneously determine the structure of the tree and learn the classifiers for each node in the tree. This approach also allows fine grained control over the efficiency vs accuracy trade-off in designing a label tree, leading to more balanced trees. Experiments are performed on large scale image classification with 10184 classes and 9 million images. We demonstrate significant improvements in test accuracy and efficiency with less training time and more balanced trees compared to the previous state of the art by Bengio et al.

The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called “ImageNet”, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500–1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond.

We present a system for semi-automatic creation of bas-relief sculpture. As an artistic medium, relief spans the continuum between 2D drawing or painting and full 3D sculpture. Bas-relief (or low relief) presents the unique challenge of squeezing shapes into a nearly-flat surface while maintaining as much as possible the perception of the full 3D scene. Our solution to this problem adapts methods from the tone-mapping literature, which addresses the similar problem of squeezing a high dynamic range image into the (low) dynamic range available on typical display devices. However, the bas-relief medium imposes its own unique set of requirements, such as maintaining small, fixed-size depth discontinuities. Given a 3D model, camera, and a few parameters describing the relative attenuation of different frequencies in the shape, our system creates a relief that gives the illusion of the 3D shape from a given vantage point while conforming to a greatly compressed height.