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.

Recent advances in generative models have sparked exciting new possibilities in the field of autonomous vehicles. Specifically, video generation models are now being explored as controllable virtual testing environments. Simultaneously, end-to-end (E2E) driving models have emerged as a streamlined alternative to conventional modular autonomous driving systems, gaining popularity for their simplicity and scalability. However, the application of these techniques to simulation and planning raises important questions. First, while video generation models can generate increasingly realistic videos, can these videos faithfully adhere to the specified conditions and be realistic enough for E2E autonomous planner evaluation? Second, given that data is crucial for understanding and controlling E2E planners, how can we gain deeper insights into their biases and improve their ability to generalize to out-of-distribution scenarios? In this work, we bridge the gap between the driving models and generative world models (Drive&Gen) to address these questions. We propose novel statistical measures leveraging E2E drivers to evaluate the realism of generated videos. By exploiting the controllability of the video generation model, we conduct targeted experiments to investigate distribution gaps affecting E2E planner performance. Finally, we show that synthetic data produced by the video generation model offers a cost-effective alternative to real-world data collection. This synthetic data effectively improves E2E model generalization beyond existing Operational Design Domains, facilitating the expansion of autonomous vehicle services into new operational contexts.

The 3D Average Precision (3DAP) relies on the intersection over union between predictions and ground truth objects. However, camera-only detectors have limited depth accuracy, which may cause otherwise reasonable predictions that suffer from such longitudinal localization errors to be treated as false positives. We therefore propose variants of the 3DAP metric to be more permissive with respect to depth estimation errors. Specifically, our novel longitudinal error tolerant metrics, LET-3D-AP and LET-3D-APL, allow longitudinal localization errors of the prediction boxes up to a given tolerance. To evaluate the proposed metrics, we also construct a new test set for the Waymo Open Dataset, tailored to camera-only 3D detection methods. Surprisingly, we find that state-of-the-art camera-based detectors can outperform popular LiDAR-based detectors with our new metrics past at 10% depth error tolerance, suggesting that existing camera-based detectors already have the potential to surpass LiDAR-based detectors in downstream applications. We believe the proposed metrics and the new benchmark dataset will facilitate advances in the field of camera-only 3D detection by providing more informative signals that can better indicate the system-level performance.

Unsupervised semantic segmentation aims to discover groupings within and across images that capture object-and view-invariance of a category without external supervision. Grouping naturally has levels of granularity, creating ambiguity in unsupervised segmentation. Existing methods avoid this ambiguity and treat it as a factor outside modeling, whereas we embrace it and desire hierarchical grouping consistency for unsupervised segmentation. We approach unsupervised segmentation as a pixel-wise feature learning problem. Our idea is that a good representation shall reveal not just a particular level of grouping, but any level of grouping in a consistent and predictable manner. We enforce spatial consistency of grouping and bootstrap feature learning with co-segmentation among multiple views of the same image, and enforce semantic consistency across the grouping hierarchy with clustering transformers between coarse- and fine-grained features. We deliver the first data-driven unsupervised hierarchical semantic segmentation method called Hierarchical Segment Grouping (HSG). Capturing visual similarity and statistical co-occurrences, HSG also outperforms existing un-supervised segmentation methods by a large margin on five major object- and scene-centric benchmarks.

Pixel-wise losses, i.e., cross-entropy or L2, have been widely used in structured prediction tasks as a spatial extension of generic image classification or regression. However, its i.i.d. assumption neglects the structural regularity present in natural images. Various attempts have been made to incorporate structural reasoning mostly through structure priors in a cooperative way where co-occurring patterns are encouraged. We, on the other hand, approach this problem from an opposing angle and propose a new framework, Adversarial Structure Matching (ASM), for training such structured prediction networks via an adversarial process, in which we train a structure analyzer that provides the supervisory signals, the ASM loss. The structure analyzer is trained to maximize ASM loss, or to emphasize recurring multi-scale hard negative structural mistakes usually among co-occurring patterns. On the contrary, the structured prediction network is trained to reduce those mistakes and is thus enabled to distinguish fine-grained structures. As a result, training structured prediction networks using ASM reduces contextual confusion among objects and improves boundary localization. We demonstrate that ASM outperforms its pixel-wise counterpart and commonly used structure priors, GAN, on three different structured prediction tasks, namely, semantic segmentation, monocular depth estimation, and surface normal prediction.

Almost all existing deep learning approaches for semantic segmentation tackle this task as a pixel-wise classification problem. Yet humans understand a scene not in terms of pixels, but by decomposing it into perceptual groups and structures that are the basic building blocks of recognition. This motivates us to propose an end-to-end pixel-wise metric learning approach that mimics this process. In our approach, the optimal visual representation determines the right segmentation within individual images and associates segments with the same semantic classes across images. The core visual learning problem is therefore to maximize the similarity within segments and minimize the similarity between segments. Given a model trained this way, inference is performed consistently by extracting pixel-wise embeddings and clustering, with the semantic label determined by the majority vote of its nearest neighbors from an annotated set. As a result, we present the SegSort, as a first attempt using deep learning for unsupervised semantic segmentation, achieving 76% performance of its supervised counterpart. When supervision is available, SegSort shows consistent improvements over conventional approaches based on pixel-wise softmax training. Additionally, our approach produces more precise boundaries and consistent region predictions. The proposed SegSort further produces an interpretable result, as each choice of label can be easily understood from the retrieved nearest segments.

We presents a method for future localization: to predict plausible future trajectories of ego-motion in egocentric stereo images. Our paths avoid obstacles, move between objects, even turn around a corner into space behind objects. As a byproduct of the predicted trajectories, we discover the empty space occluded by foreground objects. One key innovation is the creation of an EgoRetinal map, akin to an illustrated tourist map, that 'rearranges' pixels taking into accounts depth information, the ground plane, and body motion direction, so that it allows motion planning and perception of objects on one image space. We learn to plan trajectories directly on this EgoRetinal map using first person experience of walking around in a variety of scenes. In a testing phase, given an novel scene, we find multiple hypotheses of future trajectories from the learned experience. We refine them by minimizing a cost function that describes compatibility between the obstacles in the EgoRetinal map and trajectories. We quantitatively evaluate our method to show predictive validity and apply to various real world daily activities including walking, shopping, and social interactions.

A first-person video can generate powerful physical sensations of action in an observer. In this paper, we focus on a problem of Force from Motion - decoding the sensation of 1) passive forces such as the gravity, 2) the physical scale of the motion (speed) and space, and 3) active forces exerted by the observer such as pedaling a bike or banking on a ski turn. The sensation of gravity can be observed in a natural image. We learn this image cue for predicting a gravity direction in a 2D image and integrate the prediction across images to estimate the 3D gravity direction using structure from motion. The sense of physical scale is revealed to us when the body is in a dynamically balanced state. We compute the unknown physical scale of 3D reconstructed camera motion by leveraging the torque equilibrium at a banked turn that relates the centripetal force, gravity, and the body leaning angle. The active force and torque governs 3D egomotion through the physics of rigid body dynamics. Using an inverse dynamics optimization, we directly minimize 2D reprojection error (in video) with respect to 3D world structure, active forces, and additional passive forces such as air drag and friction force. We use structure from motion with the physical scale and gravity direction as an initialization of our bundle adjustment for force estimation. Our method shows quantitatively equivalent reconstruction comparing to IMU measurements in terms of gravity and scale recovery and outperforms method based on 2D optical flow for an active action recognition task. We apply our method to first person videos of mountain biking, urban bike racing, skiing, speedflying with parachute, and wingsuit flying where inertial measurements are not accessible.