Yihan Hu

Yihan Hu, Jiazhi Yang, Li Chen et al.

Modern autonomous driving system is characterized as modular tasks in sequential order, i.e., perception, prediction, and planning. In order to perform a wide diversity of tasks and achieve advanced-level intelligence, contemporary approaches either deploy standalone models for individual tasks, or design a multi-task paradigm with separate heads. However, they might suffer from accumulative errors or deficient task coordination. Instead, we argue that a favorable framework should be devised and optimized in pursuit of the ultimate goal, i.e., planning of the self-driving car. Oriented at this, we revisit the key components within perception and prediction, and prioritize the tasks such that all these tasks contribute to planning. We introduce Unified Autonomous Driving (UniAD), a comprehensive framework up-to-date that incorporates full-stack driving tasks in one network. It is exquisitely devised to leverage advantages of each module, and provide complementary feature abstractions for agent interaction from a global perspective. Tasks are communicated with unified query interfaces to facilitate each other toward planning. We instantiate UniAD on the challenging nuScenes benchmark. With extensive ablations, the effectiveness of using such a philosophy is proven by substantially outperforming previous state-of-the-arts in all aspects. Code and models are public.

Tianshuo Xu, Yichen Xie, Depu Meng et al.

Video world models should maintain evolving states when evidence is unobserved, yet current generators often freeze hidden states upon interruption. This is not simply a capacity problem: pretrained video diffusion transformers already possess KV-cache mechanisms capable of non-local retrieval, but they are rarely trained to use them as dynamic memory. We introduce ReMind, a framework eliciting dynamic memory behavior via memory-oriented data, event-aware training, and cache adaptation. Organized around a taxonomy of 100+ dynamic events, we build a camera-annotated training mixture combining VLM-filtered real videos, generated hard dynamics, synthetic camera loops, and memory-interruption augmentations. Each clip is converted into a frame graph with protected anchors, degraded intervals, and explicit temporal gaps. A node-structured curriculum, including node-drop, noisy memory, frontier continuation, and reference-cache training, forces the model to retrieve relevant past states across interruptions rather than relying solely on local continuity. PM-RoPE, an elegant camera-phase RoPE extension, unlocks spatiotemporal retrieval at a single-attention cost while preserving pretrained pathways. ReMind achieves the best overall scores on STEVO-Bench and recovery tasks. Furthermore, general image-to-video evaluations confirm this curriculum avoids catastrophic forgetting. We will open-source our code, data, and models.

Yihan Hu, Siqi Chai, Zhening Yang et al.

As autonomous driving systems being deployed to millions of vehicles, there is a pressing need of improving the system's scalability, safety and reducing the engineering cost. A realistic, scalable, and practical simulator of the driving world is highly desired. In this paper, we present an efficient solution based on generative models which learns the dynamics of the driving scenes. With this model, we can not only simulate the diverse futures of a given driving scenario but also generate a variety of driving scenarios conditioned on various prompts. Our innovative design allows the model to operate in both full-Autoregressive and partial-Autoregressive modes, significantly improving inference and training speed without sacrificing generative capability. This efficiency makes it ideal for being used as an online reactive environment for reinforcement learning, an evaluator for planning policies, and a high-fidelity simulator for testing. We evaluated our model against two real-world datasets: the Waymo motion dataset and the nuPlan dataset. On the simulation realism and scene generation benchmark, our model achieves the state-of-the-art performance. And in the planning benchmarks, our planner outperforms the prior arts. We conclude that the proposed generative model may serve as a foundation for a variety of motion planning tasks, including data generation, simulation, planning, and online training. Source code is public at https://github.com/HorizonRobotics/GUMP/

Yichen Xie, Chensheng Peng, Mazen Abdelfattah et al. · berkeley

World foundation models aim to simulate the evolution of the real world with physically plausible behavior. Unlike prior methods that handle spatial and temporal correlations separately, we propose RAYNOVA, a geometry-agonistic multiview world model for driving scenarios that employs a dual-causal autoregressive framework. It follows both scale-wise and temporal topological orders in the autoregressive process, and leverages global attention for unified 4D spatio-temporal reasoning. Different from existing works that impose strong 3D geometric priors, RAYNOVA constructs an isotropic spatio-temporal representation across views, frames, and scales based on relative Plücker-ray positional encoding, enabling robust generalization to diverse camera setups and ego motions. We further introduce a recurrent training paradigm to alleviate distribution drift in long-horizon video generation. RAYNOVA achieves state-of-the-art multi-view video generation results on nuScenes, while offering higher throughput and strong controllability under diverse input conditions, generalizing to novel views and camera configurations without explicit 3D scene representation. Our code will be released at https://raynova-ai.github.io/.

Yihan Hu, Kun Li, Pingyuan Liang et al.

This paper presents our 2nd place solution for the NuPlan Challenge 2023. Autonomous driving in real-world scenarios is highly complex and uncertain. Achieving safe planning in the complex multimodal scenarios is a highly challenging task. Our approach, Imitation with Spatial-Temporal Heatmap, adopts the learning form of behavior cloning, innovatively predicts the future multimodal states with a heatmap representation, and uses trajectory refinement techniques to ensure final safety. The experiment shows that our method effectively balances the vehicle's progress and safety, generating safe and comfortable trajectories. In the NuPlan competition, we achieved the second highest overall score, while obtained the best scores in the ego progress and comfort metrics.

Chensheng Peng, Quentin Herau, Jiezhi Yang et al. · berkeley

We present UniQueR, a unified query-based feedforward framework for efficient and accurate 3D reconstruction from unposed images. Existing feedforward models such as DUSt3R, VGGT, and AnySplat typically predict per-pixel point maps or pixel-aligned Gaussians, which remain fundamentally 2.5D and limited to visible surfaces. In contrast, UniQueR formulates reconstruction as a sparse 3D query inference problem. Our model learns a compact set of 3D anchor points that act as explicit geometric queries, enabling the network to infer scene structure, including geometry in occluded regions--in a single forward pass. Each query encodes spatial and appearance priors directly in global 3D space (instead of per-frame camera space) and spawns a set of 3D Gaussians for differentiable rendering. By leveraging unified query interactions across multi-view features and a decoupled cross-attention design, UniQueR achieves strong geometric expressiveness while substantially reducing memory and computational cost. Experiments on Mip-NeRF 360 and VR-NeRF demonstrate that UniQueR surpasses state-of-the-art feedforward methods in both rendering quality and geometric accuracy, using an order of magnitude fewer primitives than dense alternatives.

Yu Luo, Shuo Han, Yihan Hu et al.

Standard on-policy reinforcement learning relies on heuristic clipping to enforce trust regions, but this mechanism imposes a severe cost by indiscriminately truncating high-return yet high-divergence updates. We demonstrate that explicitly constraining the policy ratio variance provides a principled local approximation to trust-region constraints, eliminating the need for binary hard clipping. By acting as a distributional ``soft brake'', this approach preserves critical gradient signals from novel discoveries while naturally down-weighting and enabling the reuse of stale, off-policy data. We introduce ${\bf R}^2{\bf VPO}$ (Ratio-Variance Regularized Policy Optimization), which implements this constraint via a primal-dual optimization framework. Extensive evaluations across $7$ LLM scales, spanning both fast and slow reasoning paradigms, and $10$ robotic control tasks demonstrate the generality of the proposed approach. R$^2$VPO achieves substantial performance gains on mathematical reasoning benchmarks, with particularly pronounced improvements on smaller models, while significantly improving sample efficiency. Furthermore, it consistently outperforms PPO baselines in continuous control domains, particularly in sparse-reward and dynamic environments. Together, these findings establish ratio-variance regularization as a principled foundation for stable and data-efficient policy optimization.

Ishaan Rawal, Shubh Gupta, Yihan Hu et al.

Vision-Language-Action (VLA) models are advancing autonomous driving by replacing modular pipelines with unified end-to-end architectures. However, current VLAs face two expensive requirements: (1) massive dataset collection, and (2) dense reasoning annotations. In this work, we address both challenges with NORD (No Reasoning for Driving). Compared to existing VLAs, NORD achieves competitive performance while being fine-tuned on <60% of the data and no reasoning annotations, resulting in 3x fewer tokens. We identify that standard Group Relative Policy Optimization (GRPO) fails to yield significant improvements when applied to policies trained on such small, reasoning-free datasets. We show that this limitation stems from difficulty bias, which disproportionately penalizes reward signals from scenarios that produce high-variance rollouts within GRPO. NORD overcomes this by incorporating Dr. GRPO, a recent algorithm designed to mitigate difficulty bias in LLMs. As a result, NORD achieves competitive performance on Waymo and NAVSIM with a fraction of the training data and no reasoning overhead, enabling more efficient autonomous systems. Website: https://nord-vla-ai.github.io/

Yihan Hu, Wenxin Shao, Bo Jiang et al.

In this report, we introduce our solution to the Occupancy and Flow Prediction challenge in the Waymo Open Dataset Challenges at CVPR 2022, which ranks 1st on the leaderboard. We have developed a novel hierarchical spatial-temporal network featured with spatial-temporal encoders, a multi-scale aggregator enriched with latent variables, and a recursive hierarchical 3D decoder. We use multiple losses including focal loss and modified flow trace loss to efficiently guide the training process. Our method achieves a Flow-Grounded Occupancy AUC of 0.8389 and outperforms all the other teams on the leaderboard.

Bo Jiang, Depu Meng, Yihan Hu et al.

Modern video generators produce visually compelling clips but still struggle with physical and motion consistency, limiting their use as reliable world simulators. Existing remedies often rely on external simulators, teacher models, or curated physics-focused data. We explore a complementary self-supervised direction: extracting motion cues from the unlabeled videos already used to train video diffusion models. We propose LaMo, which formulates a latent motion prior over frame-to-frame latent changes conditioned on the current latent and prompt. This prior is exposed through two lightweight readouts: a macro motion drift used during training as a Motion Drift Loss, and a learned micro motion field used during sampling as Motion Prior Guidance. Both components are plug-and-play with existing video diffusion backbones, requiring no architectural or I/O changes. On VideoPhy and VideoPhy2, LaMo improves CogVideoX backbones and outperforms recent physics-aware baselines that use external supervision. On VBench, it preserves overall generation quality while improving motion-related dimensions. These results suggest that unlabeled video contains useful motion supervision for improving physical fidelity in modern video diffusion models.

Haichen He, Jiayi Zhou, Sifeng Shang et al.

Real-world long video understanding requires models to perform continuous tracking, information integration and memory retention over massive temporal spans within extreme video durations. Mastering this intense cognitive load constitutes the fundamental bottleneck in long video understanding. While existing benchmarks have driven progress by scaling up video duration, their evaluation tasks often require comprehending only short and isolated video segments, falling short of capturing the challenge of ultra-long-context reasoning. To measure this cognitive load, we emphasize continuous certificate length, defined as the video length a human must continuously watch to definitively answer a given question. Driven by this metric, we introduce VideoOdyssey, a benchmark specifically designed for ultra-long-context and omni-modal video understanding. VideoOdyssey is characterized by three key features: 1) Extreme video duration and diversity: spanning 11 domains and 54 subcategories with an average video duration of 109 minutes; 2) Comprehensive evaluation scenarios: offering two subsets to address different research focuses, i.e., VideoOdyssey-V for probing the limits of visual understanding in MLLMs, and VideoOdyssey-AV for evaluating synchronized audio-visual understanding for omni-modal models; 3) Ultra-long and multi-level continuous certificates: extending the average continuous certificate to 16 minutes for VideoOdyssey-V and 12.8 minutes for VideoOdyssey-AV. Crucially, we design 5 granular levels from seconds to hours, providing a comprehensive diagnostic tool to evaluate models across varying context lengths and cognitive loads. Extensive evaluations show that bottlenecks of current MLLMs extend beyond simple retrieval to include struggles with continuous reasoning across varying context lengths, fine-grained perception, and non-verbal omni-modal understanding.

Yichen Xie, Depu Meng, Chensheng Peng et al.

Relative position embedding has become a standard mechanism for encoding positional information in Transformers. However, existing formulations are typically limited to a fixed geometric space, namely 1D sequences or regular 2D/3D grids, which restricts their applicability to many computer vision tasks that require geometric reasoning across camera views or between 2D and 3D spaces. To address this limitation, we propose URoPE, a universal extension of Rotary Position Embedding (RoPE) to cross-view or cross-dimensional geometric spaces. For each key/value image patch, URoPE samples 3D points along the corresponding camera ray at predefined depth anchors and projects them into the query image plane. Standard 2D RoPE can then be applied using the projected pixel coordinates. URoPE is a parameter-free and intrinsics-aware relative position embedding that is invariant to the choice of global coordinate systems, while remaining fully compatible with existing RoPE-optimized attention kernels. We evaluate URoPE as a plug-in positional encoding for transformer architectures across a diverse set of tasks, including novel view synthesis, 3D object detection, object tracking, and depth estimation, covering 2D-2D, 2D-3D, and temporal scenarios. Experiments show that URoPE consistently improves the performance of transformer-based models across all tasks, demonstrating its effectiveness and generality for geometric reasoning. Our project website is: https://urope-pe.github.io/.

Xuefei Wang, Jialu Wang, Fengbo Zhang et al.

Multi-agent systems (MAS) powered by large language models (LLMs) have emerged as a powerful paradigm for complex problem solving, where performance critically depends on the underlying inter-agent communication topology. However, existing topology generation methods mainly optimize for isolated tasks, while real-world deployments involve streams of evolving tasks, requiring previously effective collaboration patterns to be retained and reused rather than rediscovered or overwritten. We identify a previously underexplored failure mode, \emph{topology forgetting}, in which adapting to new tasks shifts the topology generator away from communication structures required by earlier tasks. This issue stems from cross-task misalignment in both agent-level functional semantics and relational communication structures. To address this challenge, we propose \textbf{\textsc{MasFACT}}, a geometry-aware posterior transfer framework that preserves and reuses historical collaboration knowledge as transferable topology priors. We transfer these priors across task-specific agent spaces through Fused Gromov-Wasserstein optimal transport and perform PAC-Bayes-guided conservative posterior adaptation to balance task-specific plasticity with structural stability. Experiments across class-, domain-, and task-level continual settings demonstrate that \textsc{MasFACT} consistently improves average accuracy while reducing topology forgetting compared to strong topology generation and replay-based baselines, and can be seamlessly integrated with different MAS topology generators.

Shifang Zhao, Yihan Hu, Ying Shan et al.

Editing the video content with audio alignment forms a digital human-made art in current social media. However, the time-consuming and repetitive nature of manual video editing has long been a challenge for filmmakers and professional content creators alike. In this paper, we introduce CutClaw, an autonomous multi-agent framework designed to edit hours-long raw footage into meaningful short videos that leverages the capabilities of multiple Multimodal Language Models~(MLLMs) as an agent system. It produces videos with synchronized music, followed by instructions, and a visually appealing appearance. In detail, our approach begins by employing a hierarchical multimodal decomposition that captures both fine-grained details and global structures across visual and audio footage. Then, to ensure narrative consistency, a Playwriter Agent orchestrates the whole storytelling flow and structures the long-term narrative, anchoring visual scenes to musical shifts. Finally, to construct a short edited video, Editor and Reviewer Agents collaboratively optimize the final cut via selecting fine-grained visual content based on rigorous aesthetic and semantic criteria. We conduct detailed experiments to demonstrate that CutClaw significantly outperforms state-of-the-art baselines in generating high-quality, rhythm-aligned videos. The code is available at: https://github.com/GVCLab/CutClaw.

Quentin Herau, Tianshuo Xu, Depu Meng et al.

Feed-forward 3D Gaussian Splatting methods have achieved impressive reconstruction quality for autonomous driving scenes, yet they entangle scene geometry with transient appearance properties such as lighting, weather, and time of day. This coupling prevents relighting, appearance transfer, and consistent rendering across multi-traversal data captured under varying environmental conditions. We present SpectralSplat, a method that disentangles appearance from geometry within a feed-forward Gaussian Splatting framework. Our key insight is to factor color prediction into an appearance-agnostic base stream and and appearance-conditioned adapted stream, both produced by a shared MLP conditioned on a global appearance embedding derived from DINOv2 features. To enforce disentanglement, we train with paired observations generated by a hybrid relighting pipeline that combines physics-based intrinsic decomposition with diffusion based generative refinement, and supervise with complementary consistency, reconstruction, cross-appearance, and base color losses. We further introduce an appearance-adaptable temporal history that stores appearance-agnostic features, enabling accumulated Gaussians to be re-rendered under arbitrary target appearances. Experiments demonstrate that SpectralSplat preserves the reconstruction quality of the underlying backbone while enabling controllable appearance transfer and temporally consistent relighting across driving sequences.

Siyuan Jiang, Yihan Hu, Wenjie Li et al.

Functional data, representing curves or trajectories, are ubiquitous in fields like biomedicine and motion analysis. A fundamental challenge is phase variability -- temporal misalignments that obscure underlying patterns and degrade model performance. Current methods often address registration (alignment) and classification as separate, sequential tasks. This paper introduces DeepFRC, an end-to-end deep learning framework that jointly learns diffeomorphic warping functions and a classifier within a unified architecture. DeepFRC combines a neural deformation operator for elastic alignment, a spectral representation using Fourier basis for smooth functional embedding, and a class-aware contrastive loss that promotes both intra-class coherence and inter-class separation. We provide the first theoretical guarantees for such a joint model, proving its ability to approximate optimal warpings and establishing a data-dependent generalization bound that formally links registration fidelity to classification performance. Extensive experiments on synthetic and real-world datasets demonstrate that DeepFRC consistently outperforms state-of-the-art methods in both alignment quality and classification accuracy, while ablation studies validate the synergy of its components. DeepFRC also shows notable robustness to noise, missing data, and varying dataset scales. Code is available at https://github.com/Drivergo-93589/DeepFRC.

Yiheng Lin, Yihan Hu, Chenyi Zhang et al.

Transformer-based models have recently achieved outstanding performance in image matting. However, their application to high-resolution images remains challenging due to the quadratic complexity of global self-attention. To address this issue, we propose MEMatte, a \textbf{m}emory-\textbf{e}fficient \textbf{m}atting framework for processing high-resolution images. MEMatte incorporates a router before each global attention block, directing informative tokens to the global attention while routing other tokens to a Lightweight Token Refinement Module (LTRM). Specifically, the router employs a local-global strategy to predict the routing probability of each token, and the LTRM utilizes efficient modules to simulate global attention. Additionally, we introduce a Batch-constrained Adaptive Token Routing (BATR) mechanism, which allows each router to dynamically route tokens based on image content and the stages of attention block in the network. Furthermore, we construct an ultra high-resolution image matting dataset, UHR-395, comprising 35,500 training images and 1,000 test images, with an average resolution of $4872\times6017$. This dataset is created by compositing 395 different alpha mattes across 11 categories onto various backgrounds, all with high-quality manual annotation. Extensive experiments demonstrate that MEMatte outperforms existing methods on both high-resolution and real-world datasets, significantly reducing memory usage by approximately 88% and latency by 50% on the Composition-1K benchmark. Our code is available at https://github.com/linyiheng123/MEMatte.

Matthew Strong, Wei-Jer Chang, Quentin Herau et al.

Ego-centric driving videos available online provide an abundant source of visual data for autonomous driving, yet their lack of annotations makes it difficult to learn representations that capture both semantic structure and 3D geometry. Recent advances in large feedforward spatial models demonstrate that point maps and ego-motion can be inferred in a single forward pass, suggesting a promising direction for scalable driving perception. We therefore propose a label-free, teacher-guided framework for learning autonomous driving representations directly from unposed videos. Unlike prior self-supervised approaches that focus primarily on frame-to-frame consistency, we posit that safe and reactive driving depends critically on temporal context. To this end, we leverage a feedforward architecture equipped with a lightweight autoregressive module, trained using multi-modal supervisory signals that guide the model to jointly predict current and future point maps, camera poses, semantic segmentation, and motion masks. Multi-modal teachers provide sequence-level pseudo-supervision, enabling LFG to learn a unified pseudo-4D representation from raw YouTube videos without poses, labels, or LiDAR. The resulting encoder not only transfers effectively to downstream autonomous driving planning on the NAVSIM benchmark, surpassing multi-camera and LiDAR baselines with only a single monocular camera, but also yields strong performance when evaluated on a range of semantic, geometric, and qualitative motion prediction tasks. These geometry and motion-aware features position LFG as a compelling video-centric foundation model for autonomous driving.

Yihan Hu, Xuelin Chen, Xiaodong Cun

Existing video omnimatte methods typically rely on slow, multi-stage, or inference-time optimization pipelines that fail to fully exploit powerful generative priors, producing suboptimal decompositions. Our key insight is that, if a video inpainting model can be finetuned to remove the foreground-associated effects, then it must be inherently capable of perceiving these effects, and hence can also be finetuned for the complementary task: foreground layer decomposition with associated effects. However, although naïvely finetuning the inpainting model with LoRA applied to all blocks can produce high-quality alpha mattes, it fails to capture associated effects. Our systematic analysis reveals this arises because effect-related cues are primarily encoded in specific DiT blocks and become suppressed when LoRA is applied across all blocks. To address this, we introduce EasyOmnimatte, the first unified, end-to-end video omnimatte method. Concretely, we finetune a pretrained video inpainting diffusion model to learn dual complementary experts while keeping its original weights intact: an Effect Expert, where LoRA is applied only to effect-sensitive DiT blocks to capture the coarse structure of the foreground and associated effects, and a fully LoRA-finetuned Quality Expert learns to refine the alpha matte. During sampling, Effect Expert is used for denoising at early, high-noise steps, while Quality Expert takes over at later, low-noise steps. This design eliminates the need for two full diffusion passes, significantly reducing computational cost without compromising output quality. Ablation studies validate the effectiveness of this Dual-Expert strategy. Experiments demonstrate that EasyOmnimatte sets a new state-of-the-art for video omnimatte and enables various downstream tasks, significantly outperforming baselines in both quality and efficiency.

Yihan Hu, Jianing Peng, Yiheng Lin et al.

This paper presents a novel approach to improving text-guided image editing using diffusion-based models. Text-guided image editing task poses key challenge of precisly locate and edit the target semantic, and previous methods fall shorts in this aspect. Our method introduces a Precise Semantic Localization strategy that leverages visual and textual self-attention to enhance the cross-attention map, which can serve as a regional cues to improve editing performance. Then we propose a Dual-Level Control mechanism for incorporating regional cues at both feature and latent levels, offering fine-grained control for more precise edits. To fully compare our methods with other DiT-based approaches, we construct the RW-800 benchmark, featuring high resolution images, long descriptive texts, real-world images, and a new text editing task. Experimental results on the popular PIE-Bench and RW-800 benchmarks demonstrate the superior performance of our approach in preserving background and providing accurate edits.

Chenyi Zhang, Yihan Hu, Henghui Ding et al. · gatech

Despite significant advancements in image matting, existing models heavily depend on manually-drawn trimaps for accurate results in natural image scenarios. However, the process of obtaining trimaps is time-consuming, lacking user-friendliness and device compatibility. This reliance greatly limits the practical application of all trimap-based matting methods. To address this issue, we introduce Click2Trimap, an interactive model capable of predicting high-quality trimaps and alpha mattes with minimal user click inputs. Through analyzing real users' behavioral logic and characteristics of trimaps, we successfully propose a powerful iterative three-class training strategy and a dedicated simulation function, making Click2Trimap exhibit versatility across various scenarios. Quantitative and qualitative assessments on synthetic and real-world matting datasets demonstrate Click2Trimap's superior performance compared to all existing trimap-free matting methods. Especially, in the user study, Click2Trimap achieves high-quality trimap and matting predictions in just an average of 5 seconds per image, demonstrating its substantial practical value in real-world applications.

Wei-Jer Chang, Akshay Rangesh, Kevin Joseph et al.

Developing autonomous vehicles (AVs) requires not only safety and efficiency, but also realistic, human-like behaviors that are socially aware and predictable. Achieving this requires sim agent policies that are human-like, fast, and scalable in multi-agent settings. Recent progress in imitation learning with large diffusion-based or tokenized models has shown that behaviors can be captured directly from human driving data, producing realistic policies. However, these models are computationally expensive, slow during inference, and struggle to adapt in reactive, closed-loop scenarios. In contrast, self-play reinforcement learning (RL) scales efficiently and naturally captures multi-agent interactions, but it often relies on heuristics and reward shaping, and the resulting policies can diverge from human norms. We propose SPACeR, a framework that leverages a pretrained tokenized autoregressive motion model as a centralized reference policy to guide decentralized self-play. The reference model provides likelihood rewards and KL divergence, anchoring policies to the human driving distribution while preserving RL scalability. Evaluated on the Waymo Sim Agents Challenge, our method achieves competitive performance with imitation-learned policies while being up to 10x faster at inference and 50x smaller in parameter size than large generative models. In addition, we demonstrate in closed-loop ego planning evaluation tasks that our sim agents can effectively measure planner quality with fast and scalable traffic simulation, establishing a new paradigm for testing autonomous driving policies.

Jinhyung Park, Yihan Hu, Chensheng Peng et al. · berkeley

Despite the demonstrated efficiency and performance of sparse query-based representations for perception, state-of-the-art 3D occupancy prediction methods still rely on voxel-based or dense Gaussian-based 3D representations. However, dense representations are slow, and they lack flexibility in capturing the temporal dynamics of driving scenes. Distinct from prior work, we instead summarize the scene into a compact set of 3D queries which are propagated through time in an online, streaming fashion. These queries are then decoded into semantic Gaussians at each timestep. We couple our framework with a denoising rendering objective to guide the queries and their constituent Gaussians in effectively capturing scene geometry. Owing to its efficient, query-based representation, S2GO achieves state-of-the-art performance on the nuScenes and KITTI occupancy benchmarks, outperforming prior art (e.g., GaussianWorld) by 1.5 IoU with 5.9x faster inference.

Zunjie Zhu, Yan Zhao, Yihan Hu et al.

The motion capture system that supports full-body virtual representation is of key significance for virtual reality. Compared to vision-based systems, full-body pose estimation from sparse tracking signals is not limited by environmental conditions or recording range. However, previous works either face the challenge of wearing additional sensors on the pelvis and lower-body or rely on external visual sensors to obtain global positions of key joints. To improve the practicality of the technology for virtual reality applications, we estimate full-body poses using only inertial data obtained from three Inertial Measurement Unit (IMU) sensors worn on the head and wrists, thereby reducing the complexity of the hardware system. In this work, we propose a method called Progressive Inertial Poser (ProgIP) for human pose estimation, which combines neural network estimation with a human dynamics model, considers the hierarchical structure of the kinematic chain, and employs a multi-stage progressive network estimation with increased depth to reconstruct full-body motion in real time. The encoder combines Transformer Encoder and bidirectional LSTM (TE-biLSTM) to flexibly capture the temporal dependencies of the inertial sequence, while the decoder based on multi-layer perceptrons (MLPs) transforms high-dimensional features and accurately projects them onto Skinned Multi-Person Linear (SMPL) model parameters. Quantitative and qualitative experimental results on multiple public datasets show that our method outperforms state-of-the-art methods with the same inputs, and is comparable to recent works using six IMU sensors.

Yihan Hu, Yiheng Lin, Wei Wang et al.

We aim to leverage diffusion to address the challenging image matting task. However, the presence of high computational overhead and the inconsistency of noise sampling between the training and inference processes pose significant obstacles to achieving this goal. In this paper, we present DiffMatte, a solution designed to effectively overcome these challenges. First, DiffMatte decouples the decoder from the intricately coupled matting network design, involving only one lightweight decoder in the iterations of the diffusion process. With such a strategy, DiffMatte mitigates the growth of computational overhead as the number of samples increases. Second, we employ a self-aligned training strategy with uniform time intervals, ensuring a consistent noise sampling between training and inference across the entire time domain. Our DiffMatte is designed with flexibility in mind and can seamlessly integrate into various modern matting architectures. Extensive experimental results demonstrate that DiffMatte not only reaches the state-of-the-art level on the Composition-1k test set, surpassing the best methods in the past by 5% and 15% in the SAD metric and MSE metric respectively, but also show stronger generalization ability in other benchmarks.

Yihan Hu, Zhuangzhuang Ding, Runzhou Ge et al.

There have been two streams in the 3D detection from point clouds: single-stage methods and two-stage methods. While the former is more computationally efficient, the latter usually provides better detection accuracy. By carefully examining the two-stage approaches, we have found that if appropriately designed, the first stage can produce accurate box regression. In this scenario, the second stage mainly rescores the boxes such that the boxes with better localization get selected. From this observation, we have devised a single-stage anchor-free network that can fulfill these requirements. This network, named AFDetV2, extends the previous work by incorporating a self-calibrated convolution block in the backbone, a keypoint auxiliary supervision, and an IoU prediction branch in the multi-task head. As a result, the detection accuracy is drastically boosted in the single-stage. To evaluate our approach, we have conducted extensive experiments on the Waymo Open Dataset and the nuScenes Dataset. We have observed that our AFDetV2 achieves the state-of-the-art results on these two datasets, superior to all the prior arts, including both the single-stage and the two-stage 3D detectors. AFDetV2 won the 1st place in the Real-Time 3D Detection of the Waymo Open Dataset Challenge 2021. In addition, a variant of our model AFDetV2-Base was entitled the "Most Efficient Model" by the Challenge Sponsor, showing a superior computational efficiency. To demonstrate the generality of this single-stage method, we have also applied it to the first stage of the two-stage networks. Without exception, the results show that with the strengthened backbone and the rescoring approach, the second stage refinement is no longer needed.

Runzhou Ge, Zhuangzhuang Ding, Yihan Hu et al.

In this report, we introduce our winning solution to the Real-time 3D Detection and also the "Most Efficient Model" in the Waymo Open Dataset Challenges at CVPR 2021. Extended from our last year's award-winning model AFDet, we have made a handful of modifications to the base model, to improve the accuracy and at the same time to greatly reduce the latency. The modified model, named as AFDetV2, is featured with a lite 3D Feature Extractor, an improved RPN with extended receptive field and an added sub-head that produces an IoU-aware confidence score. These model enhancements, together with enriched data augmentation, stochastic weights averaging, and a GPU-based implementation of voxelization, lead to a winning accuracy of 73.12 mAPH/L2 for our AFDetV2 with a latency of 60.06 ms, and an accuracy of 72.57 mAPH/L2 for our AFDetV2-base, entitled as the "Most Efficient Model" by the challenge sponsor, with a winning latency of 55.86 ms.

Peter Gerstoft, Yihan Hu, Michael J. Bianco et al.

We present a system for localizing sound sources in a room with several ad-hoc microphone arrays. Each circular array performs direction of arrival (DOA) estimation independently using commercial software. The DOAs are fed to a fusion center, concatenated, and used to perform the localization based on two proposed methods, which require only few labeled source locations (anchor points) for training. The first proposed method is based on principal component analysis (PCA) of the observed DOA and does not require any knowledge of anchor points. The array cluster can then perform localization on a manifold defined by the PCA of concatenated DOAs over time. The second proposed method performs localization using an affine transformation between the DOA vectors and the room manifold. The PCA has fewer requirements on the training sequence, but is less robust to missing DOAs from one of the arrays. The methods are demonstrated with five IoT 8-microphone circular arrays, placed at unspecified fixed locations in an office. Both the PCA and the affine method can easily map out a rectangle based on a few anchor points with similar accuracy. The proposed methods provide a step towards monitoring activities in a smart home and require little installation effort as the array locations are not needed.

Sijia Chen, Yu Wang, Li Huang et al.

A practical autonomous driving system urges the need to reliably and accurately detect vehicles and persons. In this report, we introduce a state-of-the-art 2D object detection system for autonomous driving scenarios. Specifically, we integrate both popular two-stage detector and one-stage detector with anchor free fashion to yield a robust detection. Furthermore, we train multiple expert models and design a greedy version of the auto ensemble scheme that automatically merges detections from different models. Notably, our overall detection system achieves 70.28 L2 mAP on the Waymo Open Dataset v1.2, ranking the 2nd place in the 2D detection track of the Waymo Open Dataset Challenges.

Yu Wang, Sijia Chen, Li Huang et al.

This technical report presents the online and real-time 2D and 3D multi-object tracking (MOT) algorithms that reached the 1st places on both Waymo Open Dataset 2D tracking and 3D tracking challenges. An efficient and pragmatic online tracking-by-detection framework named HorizonMOT is proposed for camera-based 2D tracking in the image space and LiDAR-based 3D tracking in the 3D world space. Within the tracking-by-detection paradigm, our trackers leverage our high-performing detectors used in the 2D/3D detection challenges and achieved 45.13% 2D MOTA/L2 and 63.45% 3D MOTA/L2 in the 2D/3D tracking challenges.

Zhuangzhuang Ding, Yihan Hu, Runzhou Ge et al.

In this technical report, we introduce our winning solution "HorizonLiDAR3D" for the 3D detection track and the domain adaptation track in Waymo Open Dataset Challenge at CVPR 2020. Many existing 3D object detectors include prior-based anchor box design to account for different scales and aspect ratios and classes of objects, which limits its capability of generalization to a different dataset or domain and requires post-processing (e.g. Non-Maximum Suppression (NMS)). We proposed a one-stage, anchor-free and NMS-free 3D point cloud object detector AFDet, using object key-points to encode the 3D attributes, and to learn an end-to-end point cloud object detection without the need of hand-engineering or learning the anchors. AFDet serves as a strong baseline in our winning solution and significant improvements are made over this baseline during the challenges. Specifically, we design stronger networks and enhance the point cloud data using densification and point painting. To leverage camera information, we append/paint additional attributes to each point by projecting them to camera space and gathering image-based perception information. The final detection performance also benefits from model ensemble and Test-Time Augmentation (TTA) in both the 3D detection track and the domain adaptation track. Our solution achieves the 1st place with 77.11% mAPH/L2 and 69.49% mAPH/L2 respectively on the 3D detection track and the domain adaptation track.

Runzhou Ge, Zhuangzhuang Ding, Yihan Hu et al.

High-efficiency point cloud 3D object detection operated on embedded systems is important for many robotics applications including autonomous driving. Most previous works try to solve it using anchor-based detection methods which come with two drawbacks: post-processing is relatively complex and computationally expensive; tuning anchor parameters is tricky. We are the first to address these drawbacks with an anchor free and Non-Maximum Suppression free one stage detector called AFDet. The entire AFDet can be processed efficiently on a CNN accelerator or a GPU with the simplified post-processing. Without bells and whistles, our proposed AFDet performs competitively with other one stage anchor-based methods on KITTI validation set and Waymo Open Dataset validation set.