In the first half of the year, DETR set off a wave of resume papers based on Transformers. Many people joked that there are not many CV directions left for Transformer to conquer, such as detection, segmentation, depth estimation and lane line detection.

In the first half of the year, there were also good jobs such as YOLOv4 and YOLOv5.

There are also many good jobs in the second half of the year, such as the deformable DETR introduced in this paper, and the sparse R-CNN and DeFCN recently released.