Shuhan Tan

Shuhan Tan, Tushar Nagarajan, Kristen Grauman

Recent advances in egocentric video understanding models are promising, but their heavy computational expense is a barrier for many real-world applications. To address this challenge, we propose EgoDistill, a distillation-based approach that learns to reconstruct heavy egocentric video clip features by combining the semantics from a sparse set of video frames with the head motion from lightweight IMU readings. We further devise a novel self-supervised training strategy for IMU feature learning. Our method leads to significant improvements in efficiency, requiring 200x fewer GFLOPs than equivalent video models. We demonstrate its effectiveness on the Ego4D and EPICKitchens datasets, where our method outperforms state-of-the-art efficient video understanding methods.

Shuhan Tan, Boris Ivanovic, Xinshuo Weng et al.

Simulation forms the backbone of modern self-driving development. Simulators help develop, test, and improve driving systems without putting humans, vehicles, or their environment at risk. However, simulators face a major challenge: They rely on realistic, scalable, yet interesting content. While recent advances in rendering and scene reconstruction make great strides in creating static scene assets, modeling their layout, dynamics, and behaviors remains challenging. In this work, we turn to language as a source of supervision for dynamic traffic scene generation. Our model, LCTGen, combines a large language model with a transformer-based decoder architecture that selects likely map locations from a dataset of maps, and produces an initial traffic distribution, as well as the dynamics of each vehicle. LCTGen outperforms prior work in both unconditional and conditional traffic scene generation in terms of realism and fidelity. Code and video will be available at https://ariostgx.github.io/lctgen.

Shuhan Tan, Boris Ivanovic, Yuxiao Chen et al.

Simulation stands as a cornerstone for safe and efficient autonomous driving development. At its core a simulation system ought to produce realistic, reactive, and controllable traffic patterns. In this paper, we propose ProSim, a multimodal promptable closed-loop traffic simulation framework. ProSim allows the user to give a complex set of numerical, categorical or textual prompts to instruct each agent's behavior and intention. ProSim then rolls out a traffic scenario in a closed-loop manner, modeling each agent's interaction with other traffic participants. Our experiments show that ProSim achieves high prompt controllability given different user prompts, while reaching competitive performance on the Waymo Sim Agents Challenge when no prompt is given. To support research on promptable traffic simulation, we create ProSim-Instruct-520k, a multimodal prompt-scenario paired driving dataset with over 10M text prompts for over 520k real-world driving scenarios. We will release code of ProSim as well as data and labeling tools of ProSim-Instruct-520k at https://ariostgx.github.io/ProSim.

Boyi Li, Ligeng Zhu, Ran Tian et al.

We propose Wolf, a WOrLd summarization Framework for accurate video captioning. Wolf is an automated captioning framework that adopts a mixture-of-experts approach, leveraging complementary strengths of Vision Language Models (VLMs). By utilizing both image and video models, our framework captures different levels of information and summarizes them efficiently. Our approach can be applied to enhance video understanding, auto-labeling, and captioning. To evaluate caption quality, we introduce CapScore, an LLM-based metric to assess the similarity and quality of generated captions compared to the ground truth captions. We further build four human-annotated datasets in three domains: autonomous driving, general scenes, and robotics, to facilitate comprehensive comparisons. We show that Wolf achieves superior captioning performance compared to state-of-the-art approaches from the research community (VILA1.5, CogAgent) and commercial solutions (Gemini-Pro-1.5, GPT-4V). For instance, in comparison with GPT-4V, Wolf improves CapScore both quality-wise by 55.6% and similarity-wise by 77.4% on challenging driving videos. Finally, we establish a benchmark for video captioning and introduce a leaderboard, aiming to accelerate advancements in video understanding, captioning, and data alignment. Webpage: https://wolfv0.github.io/.

Shuhan Tan, Kashyap Chitta, Yuxiao Chen et al.

Recent Vision-Language-Action (VLA) models for autonomous driving explore inference-time reasoning as a way to improve driving performance and safety in challenging scenarios. Most prior work uses natural language to express chain-of-thought (CoT) reasoning before producing driving actions. However, text may not be the most efficient representation for reasoning. In this work, we present Latent-CoT-Drive (LCDrive): a model that expresses CoT in a latent language that captures possible outcomes of the driving actions being considered. Our approach unifies CoT reasoning and decision making by representing both in an action-aligned latent space. Instead of natural language, the model reasons by interleaving (1) action-proposal tokens, which use the same vocabulary as the model's output actions; and (2) world model tokens, which are grounded in a learned latent world model and express future outcomes of these actions. We cold start latent CoT by supervising the model's action proposals and world model tokens based on ground-truth future rollouts of the scene. We then post-train with closed-loop reinforcement learning to strengthen reasoning capabilities. On a large-scale end-to-end driving benchmark, LCDrive achieves faster inference, better trajectory quality, and larger improvements from interactive reinforcement learning compared to both non-reasoning and text-reasoning baselines.

Shuhan Tan, Kairan Dou, Yue Zhao et al.

We introduce RIPT-VLA, a simple and scalable reinforcement-learning-based interactive post-training paradigm that fine-tunes pretrained Vision-Language-Action (VLA) models using only sparse binary success rewards. Existing VLA training pipelines rely heavily on offline expert demonstration data and supervised imitation, limiting their ability to adapt to new tasks and environments under low-data regimes. RIPT-VLA addresses this by enabling interactive post-training with a stable policy optimization algorithm based on dynamic rollout sampling and leave-one-out advantage estimation. RIPT-VLA has the following characteristics. First, it applies to various VLA models, resulting in an improvement on the lightweight QueST model by 21.2%, and the 7B OpenVLA-OFT model to an unprecedented 97.5% success rate. Second, it is computationally efficient and data-efficient: with only one demonstration, RIPT-VLA enables an unworkable SFT model (4%) to succeed with a 97% success rate within 15 iterations. Furthermore, we demonstrate that the policy learned by RIPT-VLA generalizes across different tasks and scenarios and is robust to the initial state context. These results highlight RIPT-VLA as a practical and effective paradigm for post-training VLA models through minimal supervision.

Shuhan Tan, John Lambert, Hong Jeon et al.

The goal of traffic simulation is to augment a potentially limited amount of manually-driven miles that is available for testing and validation, with a much larger amount of simulated synthetic miles. The culmination of this vision would be a generative simulated city, where given a map of the city and an autonomous vehicle (AV) software stack, the simulator can seamlessly simulate the trip from point A to point B by populating the city around the AV and controlling all aspects of the scene, from animating the dynamic agents (e.g., vehicles, pedestrians) to controlling the traffic light states. We refer to this vision as CitySim, which requires an agglomeration of simulation technologies: scene generation to populate the initial scene, agent behavior modeling to animate the scene, occlusion reasoning, dynamic scene generation to seamlessly spawn and remove agents, and environment simulation for factors such as traffic lights. While some key technologies have been separately studied in various works, others such as dynamic scene generation and environment simulation have received less attention in the research community. We propose SceneDiffuser++, the first end-to-end generative world model trained on a single loss function capable of point A-to-B simulation on a city scale integrating all the requirements above. We demonstrate the city-scale traffic simulation capability of SceneDiffuser++ and study its superior realism under long simulation conditions. We evaluate the simulation quality on an augmented version of the Waymo Open Motion Dataset (WOMD) with larger map regions to support trip-level simulation.

Xiuyu Yang, Shuhan Tan, Philipp Krähenbühl

An ideal traffic simulator replicates the realistic long-term point-to-point trip that a self-driving system experiences during deployment. Prior models and benchmarks focus on closed-loop motion simulation for initial agents in a scene. This is problematic for long-term simulation. Agents enter and exit the scene as the ego vehicle enters new regions. We propose InfGen, a unified next-token prediction model that performs interleaved closed-loop motion simulation and scene generation. InfGen automatically switches between closed-loop motion simulation and scene generation mode. It enables stable long-term rollout simulation. InfGen performs at the state-of-the-art in short-term (9s) traffic simulation, and significantly outperforms all other methods in long-term (30s) simulation. The code and model of InfGen will be released at https://orangesodahub.github.io/InfGen

Shuhan Tan, Kelvin Wong, Shenlong Wang et al.

We consider the problem of generating realistic traffic scenes automatically. Existing methods typically insert actors into the scene according to a set of hand-crafted heuristics and are limited in their ability to model the true complexity and diversity of real traffic scenes, thus inducing a content gap between synthesized traffic scenes versus real ones. As a result, existing simulators lack the fidelity necessary to train and test self-driving vehicles. To address this limitation, we present SceneGen, a neural autoregressive model of traffic scenes that eschews the need for rules and heuristics. In particular, given the ego-vehicle state and a high definition map of surrounding area, SceneGen inserts actors of various classes into the scene and synthesizes their sizes, orientations, and velocities. We demonstrate on two large-scale datasets SceneGen's ability to faithfully model distributions of real traffic scenes. Moreover, we show that SceneGen coupled with sensor simulation can be used to train perception models that generalize to the real world.

Shuhan Tan, Yujun Shen, Bolei Zhou

Generative Adversarial Networks (GANs) advance face synthesis through learning the underlying distribution of observed data. Despite the high-quality generated faces, some minority groups can be rarely generated from the trained models due to a biased image generation process. To study the issue, we first conduct an empirical study on a pre-trained face synthesis model. We observe that after training the GAN model not only carries the biases in the training data but also amplifies them to some degree in the image generation process. To further improve the fairness of image generation, we propose an interpretable baseline method to balance the output facial attributes without retraining. The proposed method shifts the interpretable semantic distribution in the latent space for a more balanced image generation while preserving the sample diversity. Besides producing more balanced data regarding a particular attribute (e.g., race, gender, etc.), our method is generalizable to handle more than one attribute at a time and synthesize samples of fine-grained subgroups. We further show the positive applicability of the balanced data sampled from GANs to quantify the biases in other face recognition systems, like commercial face attribute classifiers and face super-resolution algorithms.

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.

Shuhan Tan, Xingchao Peng, Kate Saenko

Unsupervised domain adaptation is a promising way to generalize deep models to novel domains. However, the current literature assumes that the label distribution is domain-invariant and only aligns the feature distributions or vice versa. In this work, we explore the more realistic task of Class-imbalanced Domain Adaptation: How to align feature distributions across domains while the label distributions of the two domains are also different? Taking a practical step towards this problem, we constructed the first benchmark with 22 cross-domain tasks from 6real-image datasets. We conducted comprehensive experiments on 10 recent domain adaptation methods and find most of them are very fragile in the face of coexisting feature and label distribution shift. Towards a better solution, we further proposed a feature and label distribution CO-ALignment (COAL) model with a novel combination of existing ideas. COAL is empirically shown to outperform the most recent domain adaptation methods on our benchmarks. We believe the provided benchmarks, empirical analysis results, and the COAL baseline could stimulate and facilitate future research towards this important problem.

Shuhan Tan, Jiening Jiao, Wei-Shi Zheng

In conventional domain adaptation, a critical assumption is that there exists a fully labeled domain (source) that contains the same label space as another unlabeled or scarcely labeled domain (target). However, in the real world, there often exist application scenarios in which both domains are partially labeled and not all classes are shared between these two domains. Thus, it is meaningful to let partially labeled domains learn from each other to classify all the unlabeled samples in each domain under an open-set setting. We consider this problem as weakly supervised open-set domain adaptation. To address this practical setting, we propose the Collaborative Distribution Alignment (CDA) method, which performs knowledge transfer bilaterally and works collaboratively to classify unlabeled data and identify outlier samples. Extensive experiments on the Office benchmark and an application on person reidentification show that our method achieves state-of-the-art performance.