Wei-Chiu Ma

Benlin Liu, Yuhao Dong, Yiqin Wang et al. · deepmind, tsinghua

Multimodal language models (MLLMs) are increasingly being applied in real-world environments, necessitating their ability to interpret 3D spaces and comprehend temporal dynamics. Current methods often rely on specialized architectural designs or task-specific fine-tuning to achieve this. We introduce Coarse Correspondences, a simple lightweight method that enhances MLLMs' spatial-temporal reasoning with 2D images as input, without modifying the architecture or requiring task-specific fine-tuning. Our method uses a lightweight tracking model to identify primary object correspondences between frames in a video or across different image viewpoints, and then conveys this information to MLLMs through visual prompting. We demonstrate that this simple training-free approach brings substantial gains to GPT4-V/O consistently on four benchmarks that require spatial-temporal reasoning, including +20.5\% improvement on ScanQA, +9.7\% on OpenEQA's episodic memory subset, +6.0\% on the long-form video benchmark EgoSchema, and +11\% on the R2R navigation benchmark. Additionally, we show that Coarse Correspondences can also enhance open-source MLLMs' spatial reasoning (by +6.9\% on ScanQA) when applied in both training and inference and that the improvement can generalize to unseen datasets such as SQA3D (+3.1\%). Taken together, we show that Coarse Correspondences effectively and efficiently boosts models' performance on downstream tasks requiring spatial-temporal reasoning.

Ze Yang, Yun Chen, Jingkang Wang et al. · utoronto

Rigorously testing autonomy systems is essential for making safe self-driving vehicles (SDV) a reality. It requires one to generate safety critical scenarios beyond what can be collected safely in the world, as many scenarios happen rarely on public roads. To accurately evaluate performance, we need to test the SDV on these scenarios in closed-loop, where the SDV and other actors interact with each other at each timestep. Previously recorded driving logs provide a rich resource to build these new scenarios from, but for closed loop evaluation, we need to modify the sensor data based on the new scene configuration and the SDV's decisions, as actors might be added or removed and the trajectories of existing actors and the SDV will differ from the original log. In this paper, we present UniSim, a neural sensor simulator that takes a single recorded log captured by a sensor-equipped vehicle and converts it into a realistic closed-loop multi-sensor simulation. UniSim builds neural feature grids to reconstruct both the static background and dynamic actors in the scene, and composites them together to simulate LiDAR and camera data at new viewpoints, with actors added or removed and at new placements. To better handle extrapolated views, we incorporate learnable priors for dynamic objects, and leverage a convolutional network to complete unseen regions. Our experiments show UniSim can simulate realistic sensor data with small domain gap on downstream tasks. With UniSim, we demonstrate closed-loop evaluation of an autonomy system on safety-critical scenarios as if it were in the real world.

Yuwen Xiong, Wei-Chiu Ma, Jingkang Wang et al. · utoronto

LiDAR provides accurate geometric measurements of the 3D world. Unfortunately, dense LiDARs are very expensive and the point clouds captured by low-beam LiDAR are often sparse. To address these issues, we present UltraLiDAR, a data-driven framework for scene-level LiDAR completion, LiDAR generation, and LiDAR manipulation. The crux of UltraLiDAR is a compact, discrete representation that encodes the point cloud's geometric structure, is robust to noise, and is easy to manipulate. We show that by aligning the representation of a sparse point cloud to that of a dense point cloud, we can densify the sparse point clouds as if they were captured by a real high-density LiDAR, drastically reducing the cost. Furthermore, by learning a prior over the discrete codebook, we can generate diverse, realistic LiDAR point clouds for self-driving. We evaluate the effectiveness of UltraLiDAR on sparse-to-dense LiDAR completion and LiDAR generation. Experiments show that densifying real-world point clouds with our approach can significantly improve the performance of downstream perception systems. Compared to prior art on LiDAR generation, our approach generates much more realistic point clouds. According to A/B test, over 98.5\% of the time human participants prefer our results over those of previous methods.

Jingkang Wang, Sivabalan Manivasagam, Yun Chen et al. · utoronto

Realistic simulation is key to enabling safe and scalable development of % self-driving vehicles. A core component is simulating the sensors so that the entire autonomy system can be tested in simulation. Sensor simulation involves modeling traffic participants, such as vehicles, with high quality appearance and articulated geometry, and rendering them in real time. The self-driving industry has typically employed artists to build these assets. However, this is expensive, slow, and may not reflect reality. Instead, reconstructing assets automatically from sensor data collected in the wild would provide a better path to generating a diverse and large set with good real-world coverage. Nevertheless, current reconstruction approaches struggle on in-the-wild sensor data, due to its sparsity and noise. To tackle these issues, we present CADSim, which combines part-aware object-class priors via a small set of CAD models with differentiable rendering to automatically reconstruct vehicle geometry, including articulated wheels, with high-quality appearance. Our experiments show our method recovers more accurate shapes from sparse data compared to existing approaches. Importantly, it also trains and renders efficiently. We demonstrate our reconstructed vehicles in several applications, including accurate testing of autonomy perception systems.

Zhuoling Li, Chuanrui Zhang, Wei-Chiu Ma et al.

In recent years, transformer-based detectors have demonstrated remarkable performance in 2D visual perception tasks. However, their performance in multi-view 3D object detection remains inferior to the state-of-the-art (SOTA) of convolutional neural network based detectors. In this work, we investigate this issue from the perspective of bird's-eye-view (BEV) feature generation. Specifically, we examine the BEV feature generation method employed by the transformer-based SOTA, BEVFormer, and identify its two limitations: (i) it only generates attention weights from BEV, which precludes the use of lidar points for supervision, and (ii) it aggregates camera view features to the BEV through deformable sampling, which only selects a small subset of features and fails to exploit all information. To overcome these limitations, we propose a novel BEV feature generation method, dual-view attention, which generates attention weights from both the BEV and camera view. This method encodes all camera features into the BEV feature. By combining dual-view attention with the BEVFormer architecture, we build a new detector named VoxelFormer. Extensive experiments are conducted on the nuScenes benchmark to verify the superiority of dual-view attention and VoxelForer. We observe that even only adopting 3 encoders and 1 historical frame during training, VoxelFormer still outperforms BEVFormer significantly. When trained in the same setting, VoxelFormer can surpass BEVFormer by 4.9% NDS point. Code is available at: https://github.com/Lizhuoling/VoxelFormer-public.git.

Zitian Tang, Wenjie Ye, Wei-Chiu Ma et al.

Inferring past human motion from RGB images is challenging due to the inherent uncertainty of the prediction problem. Thermal images, on the other hand, encode traces of past human-object interactions left in the environment via thermal radiation measurement. Based on this observation, we collect the first RGB-Thermal dataset for human motion analysis, dubbed Thermal-IM. Then we develop a three-stage neural network model for accurate past human pose estimation. Comprehensive experiments show that thermal cues significantly reduce the ambiguities of this task, and the proposed model achieves remarkable performance. The dataset is available at https://github.com/ZitianTang/Thermal-IM.

Wei-Chiu Ma, Anqi Joyce Yang, Shenlong Wang et al.

Recovering the spatial layout of the cameras and the geometry of the scene from extreme-view images is a longstanding challenge in computer vision. Prevailing 3D reconstruction algorithms often adopt the image matching paradigm and presume that a portion of the scene is co-visible across images, yielding poor performance when there is little overlap among inputs. In contrast, humans can associate visible parts in one image to the corresponding invisible components in another image via prior knowledge of the shapes. Inspired by this fact, we present a novel concept called virtual correspondences (VCs). VCs are a pair of pixels from two images whose camera rays intersect in 3D. Similar to classic correspondences, VCs conform with epipolar geometry; unlike classic correspondences, VCs do not need to be co-visible across views. Therefore VCs can be established and exploited even if images do not overlap. We introduce a method to find virtual correspondences based on humans in the scene. We showcase how VCs can be seamlessly integrated with classic bundle adjustment to recover camera poses across extreme views. Experiments show that our method significantly outperforms state-of-the-art camera pose estimation methods in challenging scenarios and is comparable in the traditional densely captured setup. Our approach also unleashes the potential of multiple downstream tasks such as scene reconstruction from multi-view stereo and novel view synthesis in extreme-view scenarios.

Arhan Jain, Mingtong Zhang, Kanav Arora et al. · berkeley, gatech

A significant challenge for robot learning research is our ability to accurately measure and compare the performance of robot policies. Benchmarking in robotics is historically challenging due to the stochasticity, reproducibility, and time-consuming nature of real-world rollouts. This challenge is exacerbated for recent generalist policies, which has to be evaluated across a wide variety of scenes and tasks. Evaluation in simulation offers a scalable complement to real world evaluations, but the visual and physical domain gap between existing simulation benchmarks and the real world has made them an unreliable signal for policy improvement. Furthermore, building realistic and diverse simulated environments has traditionally required significant human effort and expertise. To bridge the gap, we introduce Policy Evaluation and Environment Reconstruction in Simulation (PolaRiS), a scalable real-to-sim framework for high-fidelity simulated robot evaluation. PolaRiS utilizes neural reconstruction methods to turn short video scans of real-world scenes into interactive simulation environments. Additionally, we develop a simple simulation data co-training recipe that bridges remaining real-to-sim gaps and enables zero-shot evaluation in unseen simulation environments. Through extensive paired evaluations between simulation and the real world, we demonstrate that PolaRiS evaluations provide a much stronger correlation to real world generalist policy performance than existing simulated benchmarks. Its simplicity also enables rapid creation of diverse simulated environments. As such, this work takes a step towards distributed and democratized evaluation for the next generation of robotic foundation models.

Guangzhao He, Rundong Luo, Wei-Chiu Ma et al.

Inverse graphics is a longstanding and highly underconstrained problem that seeks to reconstruct images as editable 3D scenes which can be rendered, relit, and manipulated. In this work, we investigate whether pretrained vision-language models (VLMs) can perform executable inverse graphics directly from a single image by reconstructing a scene as an editable Blender program, without relying on specialized 2D or 3D foundation models, differentiable rendering, or multi-view supervision. We introduce Staged Executable Inverse Graphics (SEIG), an agentic framework that reconstructs a 3D scene from a single image by progressively refining scene factors including geometry, materials, composition, and lighting directly in executable Blender code space. We evaluate our framework across diverse scenes using a range of reconstruction metrics spanning pixel-level, perceptual, and semantic fidelity. Our experiments show that staged reconstruction substantially improves reconstruction fidelity, highlighting the importance of task decomposition for executable inverse graphics with general-purpose VLMs. Finally, we showcase various downstream applications enabled by the reconstructed editable Blender scenes.

Zhi-Hao Lin, Wei-Chiu Ma, Hao-Yu Hsu et al.

We present Neural Mixtures of Planar Experts (NeurMiPs), a novel planar-based scene representation for modeling geometry and appearance. NeurMiPs leverages a collection of local planar experts in 3D space as the scene representation. Each planar expert consists of the parameters of the local rectangular shape representing geometry and a neural radiance field modeling the color and opacity. We render novel views by calculating ray-plane intersections and composite output colors and densities at intersected points to the image. NeurMiPs blends the efficiency of explicit mesh rendering and flexibility of the neural radiance field. Experiments demonstrate superior performance and speed of our proposed method, compared to other 3D representations in novel view synthesis.

Hongchi Xia, Xuan Li, Zhaoshuo Li et al.

Real-world data collection for embodied agents remains costly and unsafe, calling for scalable, realistic, and simulator-ready 3D environments. However, existing scene-generation systems often rely on rule-based or task-specific pipelines, yielding artifacts and physically invalid scenes. We present SAGE, an agentic framework that, given a user-specified embodied task (e.g., "pick up a bowl and place it on the table"), understands the intent and automatically generates simulation-ready environments at scale. The agent couples multiple generators for layout and object composition with critics that evaluate semantic plausibility, visual realism, and physical stability. Through iterative reasoning and adaptive tool selection, it self-refines the scenes until meeting user intent and physical validity. The resulting environments are realistic, diverse, and directly deployable in modern simulators for policy training. Policies trained purely on this data exhibit clear scaling trends and generalize to unseen objects and layouts, demonstrating the promise of simulation-driven scaling for embodied AI. Code, demos, and the SAGE-10k dataset can be found on the project page here: https://nvlabs.github.io/sage.

Albert J. Zhai, Kuo-Hao Zeng, Jiasen Lu et al.

The ability to learn manipulation skills by watching videos of humans has the potential to unlock a new source of highly scalable data for robot learning. Here, we tackle prehensile manipulation, in which tasks involve grasping an object before performing various post-grasp motions. Human videos offer strong signals for learning the post-grasp motions, but they are less useful for learning the prerequisite grasping behaviors, especially for robots without human-like hands. A promising way forward is to use a modular policy design, leveraging a dedicated grasp generator to produce stable grasps. However, arbitrary stable grasps are often not task-compatible, hindering the robot's ability to perform the desired downstream motion. To address this challenge, we present Perceive-Simulate-Imitate (PSI), a framework for training a modular manipulation policy using human video motion data processed by paired grasp-trajectory filtering in simulation. This simulation step extends the trajectory data with grasp suitability labels, which allows for supervised learning of task-oriented grasping capabilities. We show through real-world experiments that our framework can be used to learn precise manipulation skills efficiently without any robot data, resulting in significantly more robust performance than using a grasp generator naively.

Maximus A. Pace, Prithwish Dan, Chuanruo Ning et al.

Human videos can be recorded quickly and at scale, making them an appealing source of training data for robot learning. However, humans and robots differ fundamentally in embodiment, resulting in mismatched action execution. Direct kinematic retargeting of human hand motion can therefore produce actions that are physically infeasible for robots. Despite these low-level differences, human demonstrations provide valuable motion cues about how to manipulate and interact with objects. Our key idea is to exploit the forward diffusion process: as noise is added to actions, low-level execution differences fade while high-level task guidance is preserved. We present X-Diffusion, a principled framework for training diffusion policies that maximally leverages human data without learning dynamically infeasible motions. X-Diffusion first trains a classifier to predict whether a noisy action is executed by a human or robot. Then, a human action is incorporated into policy training only after adding sufficient noise such that the classifier cannot discern its embodiment. Actions consistent with robot execution supervise fine-grained denoising at low noise levels, while mismatched human actions provide only coarse guidance at higher noise levels. Our experiments show that naive co-training under execution mismatches degrades policy performance, while X-Diffusion consistently improves it. Across five manipulation tasks, X-Diffusion achieves a 16% higher average success rate than the best baseline. The project website is available at https://portal-cornell.github.io/X-Diffusion/.

Xingyu Fu, Yushi Hu, Bangzheng Li et al.

We introduce Blink, a new benchmark for multimodal language models (LLMs) that focuses on core visual perception abilities not found in other evaluations. Most of the Blink tasks can be solved by humans "within a blink" (e.g., relative depth estimation, visual correspondence, forensics detection, and multi-view reasoning). However, we find these perception-demanding tasks cast significant challenges for current multimodal LLMs because they resist mediation through natural language. Blink reformats 14 classic computer vision tasks into 3,807 multiple-choice questions, paired with single or multiple images and visual prompting. While humans get 95.70% accuracy on average, Blink is surprisingly challenging for existing multimodal LLMs: even the best-performing GPT-4V and Gemini achieve accuracies of 51.26% and 45.72%, only 13.17% and 7.63% higher than random guessing, indicating that such perception abilities have not "emerged" yet in recent multimodal LLMs. Our analysis also highlights that specialist CV models could solve these problems much better, suggesting potential pathways for future improvements. We believe Blink will stimulate the community to help multimodal LLMs catch up with human-level visual perception.

Jieyu Zhang, Weikai Huang, Zixian Ma et al.

Benchmarks for large multimodal language models (MLMs) now serve to simultaneously assess the general capabilities of models instead of evaluating for a specific capability. As a result, when a developer wants to identify which models to use for their application, they are overwhelmed by the number of benchmarks and remain uncertain about which benchmark's results are most reflective of their specific use case. This paper introduces Task-Me-Anything, a benchmark generation engine which produces a benchmark tailored to a user's needs. Task-Me-Anything maintains an extendable taxonomy of visual assets and can programmatically generate a vast number of task instances. Additionally, it algorithmically addresses user queries regarding MLM performance efficiently within a computational budget. It contains 113K images, 10K videos, 2K 3D object assets, over 365 object categories, 655 attributes, and 335 relationships. It can generate 750M image/video question-answering pairs, which focus on evaluating MLM perceptual capabilities. Task-Me-Anything reveals critical insights: open-source MLMs excel in object and attribute recognition but lack spatial and temporal understanding; each model exhibits unique strengths and weaknesses; larger models generally perform better, though exceptions exist; and GPT4o demonstrates challenges in recognizing rotating/moving objects and distinguishing colors.

Justin Liang, Namdar Homayounfar, Wei-Chiu Ma et al.

In this paper, we propose PolyTransform, a novel instance segmentation algorithm that produces precise, geometry-preserving masks by combining the strengths of prevailing segmentation approaches and modern polygon-based methods. In particular, we first exploit a segmentation network to generate instance masks. We then convert the masks into a set of polygons that are then fed to a deforming network that transforms the polygons such that they better fit the object boundaries. Our experiments on the challenging Cityscapes dataset show that our PolyTransform significantly improves the performance of the backbone instance segmentation network and ranks 1st on the Cityscapes test-set leaderboard. We also show impressive gains in the interactive annotation setting. We release the code at https://github.com/uber-research/PolyTransform.

Hongchi Xia, Zhi-Hao Lin, Wei-Chiu Ma et al.

Creating high-quality and interactive virtual environments, such as games and simulators, often involves complex and costly manual modeling processes. In this paper, we present Video2Game, a novel approach that automatically converts videos of real-world scenes into realistic and interactive game environments. At the heart of our system are three core components:(i) a neural radiance fields (NeRF) module that effectively captures the geometry and visual appearance of the scene; (ii) a mesh module that distills the knowledge from NeRF for faster rendering; and (iii) a physics module that models the interactions and physical dynamics among the objects. By following the carefully designed pipeline, one can construct an interactable and actionable digital replica of the real world. We benchmark our system on both indoor and large-scale outdoor scenes. We show that we can not only produce highly-realistic renderings in real-time, but also build interactive games on top.

Ava Pun, Gary Sun, Jingkang Wang et al. · cmu

Different outdoor illumination conditions drastically alter the appearance of urban scenes, and they can harm the performance of image-based robot perception systems if not seen during training. Camera simulation provides a cost-effective solution to create a large dataset of images captured under different lighting conditions. Towards this goal, we propose LightSim, a neural lighting camera simulation system that enables diverse, realistic, and controllable data generation. LightSim automatically builds lighting-aware digital twins at scale from collected raw sensor data and decomposes the scene into dynamic actors and static background with accurate geometry, appearance, and estimated scene lighting. These digital twins enable actor insertion, modification, removal, and rendering from a new viewpoint, all in a lighting-aware manner. LightSim then combines physically-based and learnable deferred rendering to perform realistic relighting of modified scenes, such as altering the sun location and modifying the shadows or changing the sun brightness, producing spatially- and temporally-consistent camera videos. Our experiments show that LightSim generates more realistic relighting results than prior work. Importantly, training perception models on data generated by LightSim can significantly improve their performance.

Cheng Sun, Jaesung Choe, Charles Loop et al. · nvidia

We propose an efficient radiance field rendering algorithm that incorporates a rasterization process on adaptive sparse voxels without neural networks or 3D Gaussians. There are two key contributions coupled with the proposed system. The first is to adaptively and explicitly allocate sparse voxels to different levels of detail within scenes, faithfully reproducing scene details with $65536^3$ grid resolution while achieving high rendering frame rates. Second, we customize a rasterizer for efficient adaptive sparse voxels rendering. We render voxels in the correct depth order by using ray direction-dependent Morton ordering, which avoids the well-known popping artifact found in Gaussian splatting. Our method improves the previous neural-free voxel model by over 4db PSNR and more than 10x FPS speedup, achieving state-of-the-art comparable novel-view synthesis results. Additionally, our voxel representation is seamlessly compatible with grid-based 3D processing techniques such as Volume Fusion, Voxel Pooling, and Marching Cubes, enabling a wide range of future extensions and applications.

Matthew Wallingford, Anand Bhattad, Aditya Kusupati et al.

Three-dimensional (3D) understanding of objects and scenes play a key role in humans' ability to interact with the world and has been an active area of research in computer vision, graphics, and robotics. Large scale synthetic and object-centric 3D datasets have shown to be effective in training models that have 3D understanding of objects. However, applying a similar approach to real-world objects and scenes is difficult due to a lack of large-scale data. Videos are a potential source for real-world 3D data, but finding diverse yet corresponding views of the same content has shown to be difficult at scale. Furthermore, standard videos come with fixed viewpoints, determined at the time of capture. This restricts the ability to access scenes from a variety of more diverse and potentially useful perspectives. We argue that large scale 360 videos can address these limitations to provide: scalable corresponding frames from diverse views. In this paper, we introduce 360-1M, a 360 video dataset, and a process for efficiently finding corresponding frames from diverse viewpoints at scale. We train our diffusion-based model, Odin, on 360-1M. Empowered by the largest real-world, multi-view dataset to date, Odin is able to freely generate novel views of real-world scenes. Unlike previous methods, Odin can move the camera through the environment, enabling the model to infer the geometry and layout of the scene. Additionally, we show improved performance on standard novel view synthesis and 3D reconstruction benchmarks.

Hongchi Xia, Entong Su, Marius Memmel et al.

Creating virtual digital replicas from real-world data unlocks significant potential across domains like gaming and robotics. In this paper, we present DRAWER, a novel framework that converts a video of a static indoor scene into a photorealistic and interactive digital environment. Our approach centers on two main contributions: (i) a reconstruction module based on a dual scene representation that reconstructs the scene with fine-grained geometric details, and (ii) an articulation module that identifies articulation types and hinge positions, reconstructs simulatable shapes and appearances and integrates them into the scene. The resulting virtual environment is photorealistic, interactive, and runs in real time, with compatibility for game engines and robotic simulation platforms. We demonstrate the potential of DRAWER by using it to automatically create an interactive game in Unreal Engine and to enable real-to-sim-to-real transfer for robotics applications.

Rundong Luo, Noah Snavely, Wei-Chiu Ma

We introduce ShadowDraw, a framework that transforms ordinary 3D objects into shadow-drawing compositional art. Given a 3D object, our system predicts scene parameters, including object pose and lighting, together with a partial line drawing, such that the cast shadow completes the drawing into a recognizable image. To this end, we optimize scene configurations to reveal meaningful shadows, employ shadow strokes to guide line drawing generation, and adopt automatic evaluation to enforce shadow-drawing coherence and visual quality. Experiments show that ShadowDraw produces compelling results across diverse inputs, from real-world scans and curated datasets to generative assets, and naturally extends to multi-object scenes, animations, and physical deployments. Our work provides a practical pipeline for creating shadow-drawing art and broadens the design space of computational visual art, bridging the gap between algorithmic design and artistic storytelling. Check out our project page https://red-fairy.github.io/ShadowDraw/ for more results and an end-to-end real-world demonstration of our pipeline!

Yen-Siang Wu, Rundong Luo, Jingsen Zhu et al.

How can we tell whether a video has been sped up or slowed down? How can we generate videos at different speeds? Although videos have been central to modern computer vision research, little attention has been paid to perceiving and controlling the passage of time. In this paper, we study time as a learnable visual concept and develop models for reasoning about and manipulating the flow of time in videos. We first exploit the multimodal cues and temporal structure naturally present in videos to learn, in a self-supervised manner, to detect speed changes and estimate playback speed. We then show that these learned temporal reasoning models enable us to curate the largest slow-motion video dataset to date from noisy in-the-wild sources. Such slow-motion footage, typically filmed by high-speed cameras, contains substantially richer temporal detail than standard videos. Using this data, we further develop models capable of temporal control, including speed-conditioned video generation, which produces motion at specified playback speed, and temporal super-resolution, which tranforms low-FPS, blurry videos into high-FPS sequences with fine-grained temporal details. Our findings highlight time as a manipulable, perceptual dimension in video learning, opening doors to temporally controllable video generation, temporal forensics detection, and potentially richer world-models that understand how events unfold over time.

Shivam Duggal, Yushi Hu, Oscar Michel et al.

Despite the unprecedented progress in the field of 3D generation, current systems still often fail to produce high-quality 3D assets that are visually appealing and geometrically and semantically consistent across multiple viewpoints. To effectively assess the quality of the generated 3D data, there is a need for a reliable 3D evaluation tool. Unfortunately, existing 3D evaluation metrics often overlook the geometric quality of generated assets or merely rely on black-box multimodal large language models for coarse assessment. In this paper, we introduce Eval3D, a fine-grained, interpretable evaluation tool that can faithfully evaluate the quality of generated 3D assets based on various distinct yet complementary criteria. Our key observation is that many desired properties of 3D generation, such as semantic and geometric consistency, can be effectively captured by measuring the consistency among various foundation models and tools. We thus leverage a diverse set of models and tools as probes to evaluate the inconsistency of generated 3D assets across different aspects. Compared to prior work, Eval3D provides pixel-wise measurement, enables accurate 3D spatial feedback, and aligns more closely with human judgments. We comprehensively evaluate existing 3D generation models using Eval3D and highlight the limitations and challenges of current models.

Prithwish Dan, Kushal Kedia, Angela Chao et al.

Human videos offer a scalable way to train robot manipulation policies, but lack the action labels needed by standard imitation learning algorithms. Existing cross-embodiment approaches try to map human motion to robot actions, but often fail when the embodiments differ significantly. We propose X-Sim, a real-to-sim-to-real framework that uses object motion as a dense and transferable signal for learning robot policies. X-Sim starts by reconstructing a photorealistic simulation from an RGBD human video and tracking object trajectories to define object-centric rewards. These rewards are used to train a reinforcement learning (RL) policy in simulation. The learned policy is then distilled into an image-conditioned diffusion policy using synthetic rollouts rendered with varied viewpoints and lighting. To transfer to the real world, X-Sim introduces an online domain adaptation technique that aligns real and simulated observations during deployment. Importantly, X-Sim does not require any robot teleoperation data. We evaluate it across 5 manipulation tasks in 2 environments and show that it: (1) improves task progress by 30% on average over hand-tracking and sim-to-real baselines, (2) matches behavior cloning with 10x less data collection time, and (3) generalizes to new camera viewpoints and test-time changes. Code and videos are available at https://portal-cornell.github.io/X-Sim/.

Tianhang Cheng, Wei-Chiu Ma, Kaiyu Guan et al.

Our world is full of identical objects (\emphe.g., cans of coke, cars of same model). These duplicates, when seen together, provide additional and strong cues for us to effectively reason about 3D. Inspired by this observation, we introduce Structure from Duplicates (SfD), a novel inverse graphics framework that reconstructs geometry, material, and illumination from a single image containing multiple identical objects. SfD begins by identifying multiple instances of an object within an image, and then jointly estimates the 6DoF pose for all instances.An inverse graphics pipeline is subsequently employed to jointly reason about the shape, material of the object, and the environment light, while adhering to the shared geometry and material constraint across instances. Our primary contributions involve utilizing object duplicates as a robust prior for single-image inverse graphics and proposing an in-plane rotation-robust Structure from Motion (SfM) formulation for joint 6-DoF object pose estimation. By leveraging multi-view cues from a single image, SfD generates more realistic and detailed 3D reconstructions, significantly outperforming existing single image reconstruction models and multi-view reconstruction approaches with a similar or greater number of observations.

Rundong Luo, Matthew Wallingford, Ali Farhadi et al. · deepmind

360° videos have emerged as a promising medium to represent our dynamic visual world. Compared to the "tunnel vision" of standard cameras, their borderless field of view offers a more complete perspective of our surroundings. While existing video models excel at producing standard videos, their ability to generate full panoramic videos remains elusive. In this paper, we investigate the task of video-to-360° generation: given a perspective video as input, our goal is to generate a full panoramic video that is consistent with the original video. Unlike conventional video generation tasks, the output's field of view is significantly larger, and the model is required to have a deep understanding of both the spatial layout of the scene and the dynamics of objects to maintain spatio-temporal consistency. To address these challenges, we first leverage the abundant 360° videos available online and develop a high-quality data filtering pipeline to curate pairwise training data. We then carefully design a series of geometry- and motion-aware operations to facilitate the learning process and improve the quality of 360° video generation. Experimental results demonstrate that our model can generate realistic and coherent 360° videos from in-the-wild perspective video. In addition, we showcase its potential applications, including video stabilization, camera viewpoint control, and interactive visual question answering.

Duong H. Le, Tuan Pham, Aniruddha Kembhavi et al.

We present Piva (Preserving Identity with Variational Score Distillation), a novel optimization-based method for editing images and 3D models based on diffusion models. Specifically, our approach is inspired by the recently proposed method for 2D image editing - Delta Denoising Score (DDS). We pinpoint the limitations in DDS for 2D and 3D editing, which causes detail loss and over-saturation. To address this, we propose an additional score distillation term that enforces identity preservation. This results in a more stable editing process, gradually optimizing NeRF models to match target prompts while retaining crucial input characteristics. We demonstrate the effectiveness of our approach in zero-shot image and neural field editing. Our method successfully alters visual attributes, adds both subtle and substantial structural elements, translates shapes, and achieves competitive results on standard 2D and 3D editing benchmarks. Additionally, our method imposes no constraints like masking or pre-training, making it compatible with a wide range of pre-trained diffusion models. This allows for versatile editing without needing neural field-to-mesh conversion, offering a more user-friendly experience.

Meng-Li Shih, Wei-Chiu Ma, Lorenzo Boyice et al.

We propose ExtraNeRF, a novel method for extrapolating the range of views handled by a Neural Radiance Field (NeRF). Our main idea is to leverage NeRFs to model scene-specific, fine-grained details, while capitalizing on diffusion models to extrapolate beyond our observed data. A key ingredient is to track visibility to determine what portions of the scene have not been observed, and focus on reconstructing those regions consistently with diffusion models. Our primary contributions include a visibility-aware diffusion-based inpainting module that is fine-tuned on the input imagery, yielding an initial NeRF with moderate quality (often blurry) inpainted regions, followed by a second diffusion model trained on the input imagery to consistently enhance, notably sharpen, the inpainted imagery from the first pass. We demonstrate high-quality results, extrapolating beyond a small number of (typically six or fewer) input views, effectively outpainting the NeRF as well as inpainting newly disoccluded regions inside the original viewing volume. We compare with related work both quantitatively and qualitatively and show significant gains over prior art.

Chen-Hsuan Lin, Wei-Chiu Ma, Antonio Torralba et al.

Neural Radiance Fields (NeRF) have recently gained a surge of interest within the computer vision community for its power to synthesize photorealistic novel views of real-world scenes. One limitation of NeRF, however, is its requirement of accurate camera poses to learn the scene representations. In this paper, we propose Bundle-Adjusting Neural Radiance Fields (BARF) for training NeRF from imperfect (or even unknown) camera poses -- the joint problem of learning neural 3D representations and registering camera frames. We establish a theoretical connection to classical image alignment and show that coarse-to-fine registration is also applicable to NeRF. Furthermore, we show that naïvely applying positional encoding in NeRF has a negative impact on registration with a synthesis-based objective. Experiments on synthetic and real-world data show that BARF can effectively optimize the neural scene representations and resolve large camera pose misalignment at the same time. This enables view synthesis and localization of video sequences from unknown camera poses, opening up new avenues for visual localization systems (e.g. SLAM) and potential applications for dense 3D mapping and reconstruction.

Wei-Chiu Ma, Shenlong Wang, Jiayuan Gu et al.

We present an efficient, effective, and generic approach towards solving inverse problems. The key idea is to leverage the feedback signal provided by the forward process and learn an iterative update model. Specifically, at each iteration, the neural network takes the feedback as input and outputs an update on the current estimation. Our approach does not have any restrictions on the forward process; it does not require any prior knowledge either. Through the feedback information, our model not only can produce accurate estimations that are coherent to the input observation but also is capable of recovering from early incorrect predictions. We verify the performance of our approach over a wide range of inverse problems, including 6-DOF pose estimation, illumination estimation, as well as inverse kinematics. Comparing to traditional optimization-based methods, we can achieve comparable or better performance while being two to three orders of magnitude faster. Compared to deep learning-based approaches, our model consistently improves the performance on all metrics. Please refer to the project page for videos, animations, supplementary materials, etc.

Shivam Duggal, Zihao Wang, Wei-Chiu Ma et al.

Reconstructing high-quality 3D objects from sparse, partial observations from a single view is of crucial importance for various applications in computer vision, robotics, and graphics. While recent neural implicit modeling methods show promising results on synthetic or dense data, they perform poorly on sparse and noisy real-world data. We discover that the limitations of a popular neural implicit model are due to lack of robust shape priors and lack of proper regularization. In this work, we demonstrate highquality in-the-wild shape reconstruction using: (i) a deep encoder as a robust-initializer of the shape latent-code; (ii) regularized test-time optimization of the latent-code; (iii) a deep discriminator as a learned high-dimensional shape prior; (iv) a novel curriculum learning strategy that allows the model to learn shape priors on synthetic data and smoothly transfer them to sparse real world data. Our approach better captures the global structure, performs well on occluded and sparse observations, and registers well with the ground-truth shape. We demonstrate superior performance over state-of-the-art 3D object reconstruction methods on two real-world datasets.

Shenlong Wang, Simon Suo, Wei-Chiu Ma et al.

Standard convolutional neural networks assume a grid structured input is available and exploit discrete convolutions as their fundamental building blocks. This limits their applicability to many real-world applications. In this paper we propose Parametric Continuous Convolution, a new learnable operator that operates over non-grid structured data. The key idea is to exploit parameterized kernel functions that span the full continuous vector space. This generalization allows us to learn over arbitrary data structures as long as their support relationship is computable. Our experiments show significant improvement over the state-of-the-art in point cloud segmentation of indoor and outdoor scenes, and lidar motion estimation of driving scenes.

Ze Yang, Shenlong Wang, Sivabalan Manivasagam et al.

Constructing and animating humans is an important component for building virtual worlds in a wide variety of applications such as virtual reality or robotics testing in simulation. As there are exponentially many variations of humans with different shape, pose and clothing, it is critical to develop methods that can automatically reconstruct and animate humans at scale from real world data. Towards this goal, we represent the pedestrian's shape, pose and skinning weights as neural implicit functions that are directly learned from data. This representation enables us to handle a wide variety of different pedestrian shapes and poses without explicitly fitting a human parametric body model, allowing us to handle a wider range of human geometries and topologies. We demonstrate the effectiveness of our approach on various datasets and show that our reconstructions outperform existing state-of-the-art methods. Furthermore, our re-animation experiments show that we can generate 3D human animations at scale from a single RGB image (and/or an optional LiDAR sweep) as input.

Namdar Homayounfar, Justin Liang, Wei-Chiu Ma et al.

Annotating videos with object segmentation masks typically involves a two stage procedure of drawing polygons per object instance for all the frames and then linking them through time. While simple, this is a very tedious, time consuming and expensive process, making the creation of accurate annotations at scale only possible for well-funded labs. What if we were able to segment an object in the full video with only a single click? This will enable video segmentation at scale with a very low budget opening the door to many applications. Towards this goal, in this paper we propose a bottom up approach where given a single click for each object in a video, we obtain the segmentation masks of these objects in the full video. In particular, we construct a correlation volume that assigns each pixel in a target frame to either one of the objects in the reference frame or the background. We then refine this correlation volume via a recurrent attention module and decode the final segmentation. To evaluate the performance, we label the popular and challenging Cityscapes dataset with video object segmentations. Results on this new CityscapesVideo dataset show that our approach outperforms all the baselines in this challenging setting.

Namdar Homayounfar, Wei-Chiu Ma, Justin Liang et al.

One of the fundamental challenges to scale self-driving is being able to create accurate high definition maps (HD maps) with low cost. Current attempts to automate this process typically focus on simple scenarios, estimate independent maps per frame or do not have the level of precision required by modern self driving vehicles. In contrast, in this paper we focus on drawing the lane boundaries of complex highways with many lanes that contain topology changes due to forks and merges. Towards this goal, we formulate the problem as inference in a directed acyclic graphical model (DAG), where the nodes of the graph encode geometric and topological properties of the local regions of the lane boundaries. Since we do not know a priori the topology of the lanes, we also infer the DAG topology (i.e., nodes and edges) for each region. We demonstrate the effectiveness of our approach on two major North American Highways in two different states and show high precision and recall as well as 89% correct topology.

Namdar Homayounfar, Wei-Chiu Ma, Shrinidhi Kowshika Lakshmikanth et al.

In this paper, we tackle the problem of online road network extraction from sparse 3D point clouds. Our method is inspired by how an annotator builds a lane graph, by first identifying how many lanes there are and then drawing each one in turn. We develop a hierarchical recurrent network that attends to initial regions of a lane boundary and traces them out completely by outputting a structured polyline. We also propose a novel differentiable loss function that measures the deviation of the edges of the ground truth polylines and their predictions. This is more suitable than distances on vertices, as there exists many ways to draw equivalent polylines. We demonstrate the effectiveness of our method on a 90 km stretch of highway, and show that we can recover the right topology 92\% of the time.

Justin Liang, Namdar Homayounfar, Wei-Chiu Ma et al.

Creating high definition maps that contain precise information of static elements of the scene is of utmost importance for enabling self driving cars to drive safely. In this paper, we tackle the problem of drivable road boundary extraction from LiDAR and camera imagery. Towards this goal, we design a structured model where a fully convolutional network obtains deep features encoding the location and direction of road boundaries and then, a convolutional recurrent network outputs a polyline representation for each one of them. Importantly, our method is fully automatic and does not require a user in the loop. We showcase the effectiveness of our method on a large North American city where we obtain perfect topology of road boundaries 99.3% of the time at a high precision and recall.

Ze Yang, Siva Manivasagam, Ming Liang et al.

Sensor simulation is a key component for testing the performance of self-driving vehicles and for data augmentation to better train perception systems. Typical approaches rely on artists to create both 3D assets and their animations to generate a new scenario. This, however, does not scale. In contrast, we propose to recover the shape and motion of pedestrians from sensor readings captured in the wild by a self-driving car driving around. Towards this goal, we formulate the problem as energy minimization in a deep structured model that exploits human shape priors, reprojection consistency with 2D poses extracted from images, and a ray-caster that encourages the reconstructed mesh to agree with the LiDAR readings. Importantly, we do not require any ground-truth 3D scans or 3D pose annotations. We then incorporate the reconstructed pedestrian assets bank in a realistic LiDAR simulation system by performing motion retargeting, and show that the simulated LiDAR data can be used to significantly reduce the amount of annotated real-world data required for visual perception tasks.

Jerry Liu, Shenlong Wang, Wei-Chiu Ma et al.

We propose a very simple and efficient video compression framework that only focuses on modeling the conditional entropy between frames. Unlike prior learning-based approaches, we reduce complexity by not performing any form of explicit transformations between frames and assume each frame is encoded with an independent state-of-the-art deep image compressor. We first show that a simple architecture modeling the entropy between the image latent codes is as competitive as other neural video compression works and video codecs while being much faster and easier to implement. We then propose a novel internal learning extension on top of this architecture that brings an additional 10% bitrate savings without trading off decoding speed. Importantly, we show that our approach outperforms H.265 and other deep learning baselines in MS-SSIM on higher bitrate UVG video, and against all video codecs on lower framerates, while being thousands of times faster in decoding than deep models utilizing an autoregressive entropy model.

Jiayuan Gu, Wei-Chiu Ma, Sivabalan Manivasagam et al.

3D shape completion for real data is important but challenging, since partial point clouds acquired by real-world sensors are usually sparse, noisy and unaligned. Different from previous methods, we address the problem of learning 3D complete shape from unaligned and real-world partial point clouds. To this end, we propose a weakly-supervised method to estimate both 3D canonical shape and 6-DoF pose for alignment, given multiple partial observations associated with the same instance. The network jointly optimizes canonical shapes and poses with multi-view geometry constraints during training, and can infer the complete shape given a single partial point cloud. Moreover, learned pose estimation can facilitate partial point cloud registration. Experiments on both synthetic and real data show that it is feasible and promising to learn 3D shape completion through large-scale data without shape and pose supervision.

Namdar Homayounfar, Yuwen Xiong, Justin Liang et al.

Obtaining precise instance segmentation masks is of high importance in many modern applications such as robotic manipulation and autonomous driving. Currently, many state of the art models are based on the Mask R-CNN framework which, while very powerful, outputs masks at low resolutions which could result in imprecise boundaries. On the other hand, classic variational methods for segmentation impose desirable global and local data and geometry constraints on the masks by optimizing an energy functional. While mathematically elegant, their direct dependence on good initialization, non-robust image cues and manual setting of hyperparameters renders them unsuitable for modern applications. We propose LevelSet R-CNN, which combines the best of both worlds by obtaining powerful feature representations that are combined in an end-to-end manner with a variational segmentation framework. We demonstrate the effectiveness of our approach on COCO and Cityscapes datasets.

Sivabalan Manivasagam, Shenlong Wang, Kelvin Wong et al.

We tackle the problem of producing realistic simulations of LiDAR point clouds, the sensor of preference for most self-driving vehicles. We argue that, by leveraging real data, we can simulate the complex world more realistically compared to employing virtual worlds built from CAD/procedural models. Towards this goal, we first build a large catalog of 3D static maps and 3D dynamic objects by driving around several cities with our self-driving fleet. We can then generate scenarios by selecting a scene from our catalog and "virtually" placing the self-driving vehicle (SDV) and a set of dynamic objects from the catalog in plausible locations in the scene. To produce realistic simulations, we develop a novel simulator that captures both the power of physics-based and learning-based simulation. We first utilize ray casting over the 3D scene and then use a deep neural network to produce deviations from the physics-based simulation, producing realistic LiDAR point clouds. We showcase LiDARsim's usefulness for perception algorithms-testing on long-tail events and end-to-end closed-loop evaluation on safety-critical scenarios.

Shivam Duggal, Shenlong Wang, Wei-Chiu Ma et al.

Our goal is to significantly speed up the runtime of current state-of-the-art stereo algorithms to enable real-time inference. Towards this goal, we developed a differentiable PatchMatch module that allows us to discard most disparities without requiring full cost volume evaluation. We then exploit this representation to learn which range to prune for each pixel. By progressively reducing the search space and effectively propagating such information, we are able to efficiently compute the cost volume for high likelihood hypotheses and achieve savings in both memory and computation. Finally, an image guided refinement module is exploited to further improve the performance. Since all our components are differentiable, the full network can be trained end-to-end. Our experiments show that our method achieves competitive results on KITTI and SceneFlow datasets while running in real-time at 62ms.

Wei-Chiu Ma, Ignacio Tartavull, Ioan Andrei Bârsan et al.

In this paper we propose a novel semantic localization algorithm that exploits multiple sensors and has precision on the order of a few centimeters. Our approach does not require detailed knowledge about the appearance of the world, and our maps require orders of magnitude less storage than maps utilized by traditional geometry- and LiDAR intensity-based localizers. This is important as self-driving cars need to operate in large environments. Towards this goal, we formulate the problem in a Bayesian filtering framework, and exploit lanes, traffic signs, as well as vehicle dynamics to localize robustly with respect to a sparse semantic map. We validate the effectiveness of our method on a new highway dataset consisting of 312km of roads. Our experiments show that the proposed approach is able to achieve 0.05m lateral accuracy and 1.12m longitudinal accuracy on average while taking up only 0.3% of the storage required by previous LiDAR intensity-based approaches.

Wei-Chiu Ma, Shenlong Wang, Rui Hu et al.

In this paper we tackle the problem of scene flow estimation in the context of self-driving. We leverage deep learning techniques as well as strong priors as in our application domain the motion of the scene can be composed by the motion of the robot and the 3D motion of the actors in the scene. We formulate the problem as energy minimization in a deep structured model, which can be solved efficiently in the GPU by unrolling a Gaussian-Newton solver. Our experiments in the challenging KITTI scene flow dataset show that we outperform the state-of-the-art by a very large margin, while being 800 times faster.

Hang Zhao, Chuang Gan, Wei-Chiu Ma et al.

Sounds originate from object motions and vibrations of surrounding air. Inspired by the fact that humans is capable of interpreting sound sources from how objects move visually, we propose a novel system that explicitly captures such motion cues for the task of sound localization and separation. Our system is composed of an end-to-end learnable model called Deep Dense Trajectory (DDT), and a curriculum learning scheme. It exploits the inherent coherence of audio-visual signals from a large quantities of unlabeled videos. Quantitative and qualitative evaluations show that comparing to previous models that rely on visual appearance cues, our motion based system improves performance in separating musical instrument sounds. Furthermore, it separates sound components from duets of the same category of instruments, a challenging problem that has not been addressed before.

Hang Chu, Wei-Chiu Ma, Kaustav Kundu et al.

We tackle the problem of using 3D information in convolutional neural networks for down-stream recognition tasks. Using depth as an additional channel alongside the RGB input has the scale variance problem present in image convolution based approaches. On the other hand, 3D convolution wastes a large amount of memory on mostly unoccupied 3D space, which consists of only the surface visible to the sensor. Instead, we propose SurfConv, which "slides" compact 2D filters along the visible 3D surface. SurfConv is formulated as a simple depth-aware multi-scale 2D convolution, through a new Data-Driven Depth Discretization (D4) scheme. We demonstrate the effectiveness of our method on indoor and outdoor 3D semantic segmentation datasets. Our method achieves state-of-the-art performance with less than 30% parameters used by the 3D convolution-based approaches.

Wei-Chiu Ma, Shenlong Wang, Marcus A. Brubaker et al.

In this paper we present a robust, efficient and affordable approach to self-localization which does not require neither GPS nor knowledge about the appearance of the world. Towards this goal, we utilize freely available cartographic maps and derive a probabilistic model that exploits semantic cues in the form of sun direction, presence of an intersection, road type, speed limit as well as the ego-car trajectory in order to produce very reliable localization results. Our experimental evaluation shows that our approach can localize much faster (in terms of driving time) with less computation and more robustly than competing approaches, which ignore semantic information.

Wei-Chiu Ma, De-An Huang, Namhoon Lee et al.

We develop predictive models of pedestrian dynamics by encoding the coupled nature of multi-pedestrian interaction using game theory, and deep learning-based visual analysis to estimate person-specific behavior parameters. Building predictive models for multi-pedestrian interactions however, is very challenging due to two reasons: (1) the dynamics of interaction are complex interdependent processes, where the predicted behavior of one pedestrian can affect the actions taken by others and (2) dynamics are variable depending on an individuals physical characteristics (e.g., an older person may walk slowly while the younger person may walk faster). To address these challenges, we (1) utilize concepts from game theory to model the interdependent decision making process of multiple pedestrians and (2) use visual classifiers to learn a mapping from pedestrian appearance to behavior parameters. We evaluate our proposed model on several public multiple pedestrian interaction video datasets. Results show that our strategic planning model explains human interactions 25% better when compared to state-of-the-art methods.