In Submission Investigating positional embedding artifacts in ViTs, and denosing them.
ICRA 2024 The first work to investigate how to leverage standard definition maps to augment lane topology reasoning.
Accurate 3D object detection is crucial to autonomous driving. Though LiDAR-based detectors have achieved impressive performance, the high cost of LiDAR sensors precludes their widespread adoption in affordable vehicles. Camera-based detectors are …
Accurate 3D object detection and understanding for self-driving cars heavily relies on LiDAR point clouds, necessitating large amounts of labeled data to train. In this work, we introduce an innovative pre-training approach, Grounded Point …
NeurIPS 2023 Discovering 3D objects with reward fine-tuning, drawing inspiration from the RL community.
For a self-driving car to operate reliably, its perceptual system must generalize to the end-user's environment -- ideally without additional annotation efforts. One potential solution is to leverage unlabeled data (e.g., unlabeled LiDAR point …
CVPR 2022 Discovering 3D objects from repeated traversals and self training.
CVPR 2021 __(Oral)__
Solving non-negative image generation for Augmented Reality.
In this paper, we address the important problem in self-driving of forecasting multi-pedestrian motion and their shared scene occupancy map, critical for safe navigation. Our contributions are two-fold. First, we advocate for predicting both the …