Conditional DETR for Fast Training Convergence

Depu Meng, Xiaokang Chen, Zejia Fan +5

The recently-developed DETR approach applies the transformer encoder and decoder architecture to object detection and achieves promising performance. In this paper, we handle the c…