Test-time adaptation (TTA) intends to adapt the pretrained model to test distributions with
only unlabeled test data streams. Most of the previous TTA methods have achieved great …

This paper presents a simple yet effective approach that improves continual test-time
adaptation (TTA) in a memory-efficient manner. TTA may primarily be conducted on edge …

This paper proposes a novel test-time adaptation strategy that adjusts the model pre-trained
on the source domain using only unlabeled online data from the target domain to alleviate …

Multiple object tracking (MOT) is a fundamental component of perception systems for
autonomous driving, and its robustness to unseen conditions is a requirement to avoid life …

Recent advances in 3D object detection (3DOD) have obtained remarkably strong results for
LiDAR-based models. In contrast, surround-view 3DOD models based on multiple camera …

Abstract Bird's-eye-view (BEV) grid is a common representation for the perception of road
components, eg, drivable area, in autonomous driving. Most existing approaches rely on …

We present DejaVu, a novel framework which leverages conditional image regeneration as
additional supervision during training to improve deep networks for dense prediction tasks …

Abstract 3D point cloud segmentation has received significant interest for its growing
applications. However, the generalization ability of models suffers in dynamic scenarios due …

This paper presents a novel framework to integrate both semantic and instance contexts for
panoptic segmentation. In existing works, it is common to use a shared backbone to extract …

Contextual information plays a core role for video semantic segmentation (VSS). This paper
summarizes contexts for VSS in two-fold: local temporal contexts (LTC) which define the …