Closed becauseofAI closed 1 year ago
DETR works the same as other methods, except for a cluster-based bipartite graph matching and cross-entropy loss.
You can refer to Section B.5. for the detailed description in the paper.
Please refer to DETR pre-training config for more pre-training details, we provide an example to pre-train and fine-tune Mask R-CNN. The DETR is similar to it, and all you need to do is replace the Mask R-CNN config with the DETR config
Close this issue due to no response for 2 days. Feel free to re-open this issue if you have more questions.
How does DETR perform alignment operations? Can I find a detailed description in the paper? Where is the code?