Will the code of TransVG++ be released？ thank you

[linhuixiao]

Will the code of TransVG++ be released？
Besides, in TransVG++, the abstract mentioned it‘s a pure-Transformer framework, while in 4.2 Deatails of the paper, it says that "use Mask R-CNN [84] in TransVG++", is it a conflict?
thank you!

[linhuixiao]

As for "use Mask R-CNN [84] in TransVG++", is it a writing mistake?

[linhuixiao]

As for "use Mask R-CNN [84] in TransVG++", is it a writing mistake?

[djiajunustc]

No, it is not a writing mistake.
We use Mask R-CNN to pre-train the ViT backbone parameters in the journal version. You could also regard the pretraining process as training a ViTDet but without the window attention strategy.