2 papers
cs.CV2024
SiamMo: Siamese Motion-Centric 3D Object Tracking
Yuxiang Yang, Yingqi Deng, Jing Zhang +2
Current 3D single object tracking methods primarily rely on the Siamese matching-based paradigm, which struggles with textureless and incomplete LiDAR point clouds. Conversely, the…
cs.CV2024
VoxelTrack: Exploring Voxel Representation for 3D Point Cloud Object Tracking
Yuxuan Lu, Jiahao Nie, Zhiwei He +2
Current LiDAR point cloud-based 3D single object tracking (SOT) methods typically rely on point-based representation network. Despite demonstrated success, such networks suffer fro…