7 papers
How Far Can 5,500 Hours of Driving Take You? A Scaling Law Analysis of Video Diffusion Models
Victor Besnier, Anh-Quan Cao, Elias Ramzi +5
Video generation for autonomous driving cannot follow the web-scale route: driving data is expensive to collect, bound by privacy requirements, and cannot be scraped at will, so mo…
Vanilla ViT for Automotive Point Cloud Semantic Segmentation
Gilles Puy, Nermin Samet, Alexandre Boulch +3
Plain Transformers have become the de-facto architecture for processing text, audio, image, and video, offering a unified backbone for multimodal learning. However, state-of-the-ar…
OccAny: Generalized Unconstrained Urban 3D Occupancy
Anh-Quan Cao, Tuan-Hung Vu
Relying on in-domain annotations and precise sensor-rig priors, existing 3D occupancy prediction methods are limited in both scalability and out-of-domain generalization. While rec…
Improving Multimodal Distillation for 3D Semantic Segmentation under Domain Shift
Björn Michele, Alexandre Boulch, Gilles Puy +3
Semantic segmentation networks trained under full supervision for one type of lidar fail to generalize to unseen lidars without intervention. To reduce the performance gap under do…
VaViM and VaVAM: Autonomous Driving through Video Generative Modeling
Florent Bartoccioni, Elias Ramzi, Victor Besnier +14
We explore the potential of large-scale generative video models for autonomous driving, introducing an open-source auto-regressive video model (VaViM) and its companion video-actio…
PPT: Pretraining with Pseudo-Labeled Trajectories for Motion Forecasting
Yihong Xu, Yuan Yin, Éloi Zablocki +3
Accurately predicting how agents move in dynamic scenes is essential for safe autonomous driving. State-of-the-art motion forecasting models rely on datasets with manually annotate…
Train Till You Drop: Towards Stable and Robust Source-free Unsupervised 3D Domain Adaptation
Björn Michele, Alexandre Boulch, Tuan-Hung Vu +3
We tackle the challenging problem of source-free unsupervised domain adaptation (SFUDA) for 3D semantic segmentation. It amounts to performing domain adaptation on an unlabeled tar…