To improve the training efficiency, we propose Deformable VisTR, leveraging spatio-temporal deformable attention module that only attends to a small fixed set of key spatio-temporal sampling points around a reference point.
In addition, VisTR is not fully end-to-end learnable in multiple video clips as it requires a hand-crafted data association to link instance tracklets between successive clips.
Most online multi-object trackers perform object detection stand-alone in a neural net without any input from tracking.
Ranked #1 on Instance Segmentation on nuScenes
This task is confronted with two challenges: how to establish the 3D correspondences from views to the BEV map and how to assemble occupancy information across views.
In the classification tree, as the number of parent class nodes are significantly less, their logits are less noisy and can be utilized to suppress the wrong/noisy logits existed in the fine-grained class nodes.
Ranked #5 on Few-Shot Object Detection on LVIS v1.0 val
State-of-the-art pedestrian detectors have performed promisingly on non-occluded pedestrians, yet they are still confronted by heavy occlusions.
Ranked #16 on Pedestrian Detection on Caltech