The latest advancements in multi-modal large language models (MLLMs) have spurred a strong renewed interest in end-to-end motion planning approaches for autonomous driving. Many end-to-end approaches rely on human annotations to learn intermediate perception and prediction tasks, while purely self-supervised approaches—which directly learn from sensor inputs to generate planning trajectories without human annotations—often underperform the state of the art. We observe a key gap in the input representation space: end-to-end approaches built on MLLMs are often pretrained with reasoning tasks in 2D image space rather than the native 3D space in which autonomous vehicles plan. To this end, we propose S4-Driver, a scalable self-supervised motion planning algorithm with spatio-temporal visual representation, based on the popular PaLI [9] multimodal large language model. S4-Driver uses a novel sparse volume strategy to seamlessly transform the strong visual representation of MLLMs from perspective view to 3D space without the need to finetune the vision encoder. This representation aggregates multi-view and multi-frame visual inputs and enables better prediction of planning trajectories in 3D space. To validate our method, we run experiments on both nuScenes and Waymo Open Motion Dataset (with in-house camera data). Results show that S4-Driver performs favorably against existing supervised multi-task approaches while requiring no human annotations. It also demonstrates great scalability when pretrained on large volumes of unannotated driving logs.

We present DeSiRe-GS, a self-supervised gaussian splatting representation, enabling effective static-dynamic decomposition and high-fidelity surface reconstruction in complex driving scenarios. Our approach employs a two-stage optimization pipeline of dynamic street Gaussians. In the first stage, we extract 2D motion masks based on the observation that 3D Gaussian Splatting inherently can reconstruct only the static regions in dynamic environments. These extracted 2D motion priors are then mapped into the Gaussian space in a differentiable manner, leveraging an efficient formulation of dynamic Gaussians in the second stage. Combined with the introduced geometric regularizations, our method are able to address the over-fitting issues caused by data sparsity in autonomous driving, reconstructing physically plausible Gaussians that align with object surfaces rather than floating in air. Furthermore, we introduce temporal cross-view consistency to ensure coherence across time and viewpoints, resulting in high-quality surface reconstruction. Comprehensive experiments demonstrate the efficiency and effectiveness of DeSiRe-GS, surpassing prior self-supervised arts and achieving accuracy comparable to methods relying on external 3D bounding box annotations. Code is available at https://github.com/chengweialan/DeSiRe-GS

Recent advancements have exploited diffusion models for the synthesis of either LiDAR point clouds or camera image data in driving scenarios. Despite their success in modeling single-modality data marginal distribution, there is an under- exploration in the mutual reliance between different modalities to describe com- plex driving scenes. To fill in this gap, we propose a novel framework, X-DRIVE, to model the joint distribution of point clouds and multi-view images via a dual- branch latent diffusion model architecture. Considering the distinct geometrical spaces of the two modalities, X-DRIVE conditions the synthesis of each modality on the corresponding local regions from the other modality, ensuring better alignment and realism. To further handle the spatial ambiguity during denoising, we design the cross-modality condition module based on epipolar lines to adaptively learn the cross-modality local correspondence. Besides, X-DRIVE allows for controllable generation through multi-level input conditions, including text, bounding box, image, and point clouds. Extensive results demonstrate the high-fidelity synthetic results of X-DRIVE for both point clouds and multi-view images, adhering to input conditions while ensuring reliable cross-modality consistency. Our code will be made publicly available at https://github.com/yichen928/X-Drive.

In the field of long-tailed recognition, the Decoupled Training paradigm has shown exceptional promise by dividing training into two stages: representation learning and classifier re-training. While previous work has tried to improve both stages simultaneously, this complicates isolating the effect of classifier re-training. Recent studies reveal that simple regularization can produce strong feature representations, highlighting the need to reassess classifier re-training methods. In this study, we revisit classifier re-training methods based on a unified feature representation and re-evaluate their performances. We propose two new metrics, Logits Magnitude and Regularized Standard Deviation, to compare the differences and similarities between various methods. Using these two newly proposed metrics, we demonstrate that when the Logits Magnitude across classes is nearly balanced, further reducing its overall value can effectively decrease errors and disturbances during training, leading to better model performance. Based on our analysis using these metrics, we observe that adjusting the logits could improve model performance, leading us to develop a simple label over-smoothing approach to adjust the logits without requiring prior knowledge of class distribution. This method softens the original one-hot labels by assigning a probability slightly higher than $\frac{1}{K}$ to the true class and slightly lower than $\frac{1}{K}$ to the other classes, where $K$ is the number of classes. Our method achieves state-of-the-art performance on various imbalanced datasets, including CIFAR100-LT, ImageNet-LT, and iNaturalist2018.

Semi-supervised 3D object detection is a common strategy employed to circumvent the challenge of manually labeling large-scale autonomous driving perception datasets. Pseudo-labeling approaches to semi-supervised learning adopt a teacher-student framework in which machine-generated pseudo-labels on a large unlabeled dataset are used in combination with a small manually-labeled dataset for training. In this work, we address the problem of improving pseudo-label quality through leveraging long- term temporal information captured in driving scenes. More specifically, we leverage pre-trained motion-forecasting models to generate object trajectories on pseudo-labeled data to further enhance the student model training. Our approach improves pseudo-label quality in two distinct manners: first, we suppress false positive pseudo-labels through establishing consistency across multiple frames of motion forecasting outputs. Second, we compensate for false negative detections by directly inserting predicted object tracks into the pseudo-labeled scene. Experiments on the nuScenes dataset demonstrate the effectiveness of our approach, improving the performance of standard semi-supervised approaches in a variety of settings.

Multi-frame temporal inputs are important for vision-based autonomous driving. Observations from different angles enable the recovery of 3 D object states from 2 D images as long as we can identify the same instance from different input frames. However, the dynamic nature of driving scenes leads to significant variance in the instance appearance and shape captured by the cameras at different time steps. To this end, we propose a novel contrastive learning algorithm, Cohere3D, to learn coherent instance representations robust to the changes of distance and perspective in a long-term temporal sequence without any human annotations. In the pretraining stage, raw point clouds from LiDAR sensors are utilized to construct the instance-wise long-term temporal correspondence, which serves as guidance for the extraction of instance-level representation from the vision-based bird's-eye-view (BEV) feature map. Cohere3D encourages consistent representation for the same instance at different frames but distinguishes between different instances. We validate the effectiveness and generalizability of our algorithm by finetuning the pretrained model across key downstream autonomous driving tasks: perception, mapping, prediction, and planning. Results show a notable improvement in both data efficiency and final performance in all these tasks.

Current autonomous driving systems rely on specialized models for perceiving and predicting motion, which demonstrate reliable performance in standard conditions. However, generalizing cost-effectively to diverse real-world scenarios remains a significant challenge. To address this, we propose Plug-and-Forecast (PnF), a plug-and-play approach that augments existing motion forecasting models with multimodal large language models (MLLMs). PnF builds on the insight that natural language provides a more effective way to describe and handle complex scenarios, enabling quick adaptation to targeted behaviors. We design prompts to extract structured scene understanding from MLLMs and distill this information into learnable embeddings to augment existing behavior prediction models. Our method leverages the zero-shot reasoning capabilities of MLLMs to achieve significant improvements in motion prediction performance, while requiring no fine-tuning—making it practical to adopt. We validate our approach on two state-of-the-art motion forecasting models using the Waymo Open Motion Dataset and the nuScenes Dataset, demonstrating consistent performance improvements across both benchmarks.

The pretraining-finetuning paradigm has become dominant in computer vision, yet strategically exploiting limited annotation budgets during finetuning remains unexplored. We introduce active finetuning-a novel task for selecting the most informative samples to annotate within this paradigm. We propose Sel4FT, a unified annotation selection framework that optimizes a parametric model in continuous feature space to identify a subset preserving the entire pool's distribution while maintaining diversity. To address distribution shifts from data augmentation, we develop Sel4FT++ with augmentation-aware selection mechanisms. We theoretically prove our approach minimizes the Earth Mover's Distance between selected subset and full data pool. Our framework eliminates iterative retraining and annotation process during selection, providing an efficient solution for real-world deployment. Extensive experiments on image classification, long-tailed recognition, and semantic segmentation demonstrate state-of-the-art performance with over $100\times$100× speedup compared to existing methods.

The pretraining-finetuning paradigm has gained widespread adoption in vision tasks and other fields. However, the finetuning phase still requires high-quality annotated samples. To overcome this challenge, the concept of active finetuning has emerged, aiming to select the most appropriate samples for model finetuning within a limited budget. Existing active learning methods struggle in this scenario due to their inherent bias in batch selection. Meanwhile, the recent active finetuning approach focuses solely on global distribution alignment but neglects the contributions of samples to local boundaries. Therefore, we propose a Bi-Level Active Finetuning framework (BiLAF) to select the samples for annotation in one shot, encompassing two stages: core sample selection for global diversity and boundary sample selection for local decision uncertainty. Without the need of ground-truth labels, our method can successfully identify pseudo-class centers, apply a novel denoising technique, and iteratively select boundary samples with designed evaluation metric. Extensive experiments provide qualitative and quantitative evidence of our method's superior efficacy, consistently outperforming the existing baselines.

Road surface conditions, especially geometry profiles, enormously affect driving performance of autonomous vehicles. Vision-based online road reconstruction promisingly captures road information in advance. Existing solutions like monocular depth estimation and stereo matching suffer from modest performance. The recent technique of Bird’s-Eye-View (BEV) perception provides immense potential to more reliable and accurate reconstruction. This paper uniformly proposes two simple yet effective models for road elevation reconstruction in BEV named RoadBEV-mono and RoadBEV-stereo, which estimate road elevation with monocular and stereo images, respectively. The former directly fits elevation values based on voxel features queried from image view, while the latter efficiently recognizes road elevation patterns based on BEV volume representing correlation between left and right voxel features. Insightful analyses reveal their consistence and difference with the perspective view. Experiments on real-world dataset verify the models’ effectiveness and superiority. Elevation errors of RoadBEV-mono and RoadBEV-stereo achieve 1.83 cm and 0.50 cm, respectively. Our models are promising for practical road preview, providing essential information for promoting safety and comfort of autonomous vehicles. The code is released athttps://github.com/ztsrxh/RoadBEV.

Given the large-scale data and the high annotation cost, pretraining-finetuning becomes a popular paradigm in multiple computer vision tasks. Previous research has covered both the unsupervised pretraining and supervised finetuning in this paradigm, while little attention is paid to exploiting the annotation budget for finetuning. To fill in this gap, we formally define this new active finetuning task focusing on the selection of samples for annotation in the pretraining-finetuning paradigm. We propose a novel method called ActiveFT for active finetuning task to select a subset of data distributing similarly with the entire unlabeled pool and maintaining enough diversity by optimizing a parametric model in the continuous space. We prove that the Earth Mover's distance between the distributions of the selected subset and the entire data pool is also reduced in this process. Extensive experiments show the leading performance and high efficiency of ActiveFT superior to baselines on both image classification and semantic segmentation. Our code is released at https://github.com/yichen928/ActiveFT.

By identifying four important components of existing LiDAR-camera 3D object detection methods (LiDAR and camera candidates, transformation, and fusion outputs), we observe that all existing methods either find dense candidates or yield dense representations of scenes. However, given that objects occupy only a small part of a scene, finding dense candidates and generating dense representations is noisy and inefficient. We propose SparseFusion, a novel multi-sensor 3D detection method that exclusively uses sparse candidates and sparse representations. Specifically, SparseFusion utilizes the outputs of parallel detectors in the LiDAR and camera modalities as sparse candidates for fusion. We transform the camera candidates into the LiDAR coordinate space by disentangling the object representations. Then, we can fuse the multi-modality candidates in a unified 3D space by a lightweight self-attention module. To mitigate negative transfer between modalities, we propose novel semantic and geometric cross-modality transfer modules that are applied prior to the modality-specific detectors. SparseFusion achieves state-of-the-art performance on the nuScenes benchmark while also running at the fastest speed, even outperforming methods with stronger backbones. We perform extensive experiments to demonstrate the effectiveness and efficiency of our modules and overall method pipeline. Our code will be made publicly available at https://github.com/yichen928/SparseFusion.

Humans are capable of abstracting various tasks as different combinations of multiple attributes. This perspective of compositionality is vital for human rapid learning and adaption since previous experiences from related tasks can be combined to generalize across novel compositional settings. In this work, we aim to achieve zero-shot policy generalization of Reinforcement Learning (RL) agents by leveraging the task compositionality. Our proposed method is a meta-RL algorithm with disentangled task representation, explicitly encoding different aspects of the tasks. Policy generalization is then performed by inferring unseen compositional task representations via the obtained disentanglement without extra exploration. The evaluation is conducted on three simulated tasks and a challenging real-world robotic insertion task. Experimental results demonstrate that our proposed method achieves policy generalization to unseen compositional tasks in a zero-shot manner.

A desirable data selection algorithm can efficiently choose the most informative samples to maximize the utility of limited annotation budgets. However, current approaches, represented by active learning methods, typically follow a cumbersome pipeline that iterates the time-consuming model training and batch data selection repeatedly. In this paper, we challenge this status quo by designing a distinct data selection pipeline that utilizes existing general-purpose models to select data from various datasets with a single-pass inference without the need for additional training or supervision. A novel free data selection (FreeSel) method is proposed following this new pipeline. Specifically, we define semantic patterns extracted from inter-mediate features of the general-purpose model to capture subtle local information in each image. We then enable the selection of all data samples in a single pass through distance-based sampling at the fine-grained semantic pattern level. FreeSel bypasses the heavy batch selection process, achieving a significant improvement in efficiency and being 530x faster than existing active learning methods. Extensive experiments verify the effectiveness of FreeSel on various computer vision tasks.

As the basis for prehensile manipulation, it is vital to enable robots to grasp as robustly as humans. Our innate grasping system is prompt, accurate, flexible, and continuous across spatial and temporal domains. Few existing methods cover all these properties for robot grasping. In this article, we propose AnyGrasp for grasp perception to enable robots these abilities using a parallel gripper. Specifically, we develop a dense supervision strategy with real perception and analytic labels in the spatial–temporal domain. Additional awareness of objects' center-of-mass is incorporated into the learning process to help improve grasping stability. Utilization of grasp correspondence across observations enables dynamic grasp tracking. Our model can efficiently generate accurate, 7-DoF, dense, and temporally-smooth grasp poses and works robustly against large depth-sensing noise. Using AnyGrasp, we achieve a 93.3% success rate when clearing bins with over 300 unseen objects, which is on par with human subjects under controlled conditions. Over 900 mean-picks-per-hour is reported on a single-arm system. For dynamic grasping, we demonstrate catching swimming robot fish in the water.

This paper aims to explain deep neural networks (DNNs) from the perspective of multivariate interactions. In this paper, we define and quantify the significance of interactions among multiple input variables of the DNN. Input variables with strong interactions usually form a coalition and reflect prototype features, which are memorized and used by the DNN for inference. We define the significance of interactions based on the Shapley value, which is designed to assign the attribution value of each input variable to the inference. We have conducted experiments with various DNNs. Experimental results have demonstrated the effectiveness of the proposed method.

Recent years, human-object interaction (HOI) detection has achieved impressive advances. However, conventional two-stage methods are usually slow in inference. On the other hand, existing one-stage methods mainly focus on the union regions of interactions, which introduce unnecessary visual information as disturbances to HOI detection. To tackle the problems above, we propose a novel one-stage HOI detection approach DIRV in this paper, based on a new concept called interaction region for the HOI problem. Unlike previous methods, our approach concentrates on the densely sampled interaction regions across different scales for each human-object pair, so as to capture the subtle visual features that is most essential to the interaction. Moreover, in order to compensate for the detection flaws of a single interaction region, we introduce a novel voting strategy that makes full use of those overlapped interaction regions in place of conventional Non-Maximal Suppression (NMS). Extensive experiments on two popular benchmarks: V-COCO and HICO-DET show that our approach outperforms existing state-of-the-arts by a large margin with the highest inference speed and lightest network architecture. Our code is publicly available at www.github.com/MVIG-SJTU/DIRV.

Human-object interaction (HOI) detection requires a large amount of annotated data. Current algorithms suffer from insufficient training samples and category imbalance within datasets. To increase data efficiency, in this paper, we propose an efficient and effective data augmentation method called DecAug for HOI detection. Based on our proposed object state similarity metric, object patterns across different HOIs are shared to augment local object appearance features without changing their states. Further, we shift spatial correlation between humans and objects to other feasible configurations with the aid of a pose-guided Gaussian Mixture Model while preserving their interactions. Experiments show that our method brings up to 3.3 mAP and 1.6 mAP improvements on V-COCO and HICO-DET dataset for two advanced models. Specifically, interactions with fewer samples enjoy more notable improvement. Our method can be easily integrated into various HOI detection models with negligible extra computational consumption.

To date, various 3D scene understanding tasks still lack practical and generalizable pre-trained models, primarily due to the intricate nature of 3D scene understanding tasks and their immense variations introduced by camera views, lighting, occlusions, etc. In this paper, we tackle this challenge by introducing a spatio-temporal representation learning (STRL) framework, capable of learning from unlabeled 3D point clouds in a self-supervised fashion. Inspired by how infants learn from visual data in the wild, we explore the rich spatio-temporal cues derived from the 3D data. Specifically, STRL takes two temporally-correlated frames from a 3D point cloud sequence as the input, transforms it with the spatial data augmentation, and learns the invariant representation self-supervisedly. To corroborate the efficacy of STRL, we conduct extensive experiments on three types (synthetic, indoor, and outdoor) of datasets. Experimental results demonstrate that, compared with supervised learning methods, the learned self-supervised representation facilitates various models to attain comparable or even better performances while capable of generalizing pre-trained models to downstream tasks, including 3D shape classification, 3D object detection, and 3D semantic segmentation. Moreover, the spatio-temporal contextual cues embedded in 3D point clouds significantly improve the learned representations.