Yuxin Fang (方羽新)
Yuxin Fang (方羽新)
Hello @AK391 and thank you for your interest in our work! We will work on this and add our MIMDet to Hugging Face soon!
Try the notebook.
> @Yuxin-CV Can you tell me more? Try https://github.com/hustvl/YOLOS/blob/main/VisualizeAttention.ipynb
Hi, for the memory issue, please refer to https://github.com/hustvl/YOLOS/issues/5#issuecomment-867533669
> For those interested, I found that the HF implementation is set up for Gradient Accumulation. > > Enable it with: > > ```python > self.model = YolosForObjectDetection.from_pretrained( > self.hparams.pretrained_model_name_or_path,...
Hi, @normster. Thanks for your interest in our work. I suggest scaling down the batch size by 4x & scaling down the lr by 2x.
I also suggest aligning your environment with ours, please also see https://github.com/SysCV/transfiner/issues/17#issuecomment-1115853393.
Hello! To my knowledge, MAE uses 2D sin-cos pos embed while YOLOS uses 1D abs learnable pos embed. I suggest changing the original YOLOS pos embed to MAE's.
Hi, we use Detectron2 for Mask R-CNN as the detector. MIMDet is essentially a adapted **backbone \ feature extractor** from MAE pre-trained representations. So in principal it is ok to...
Our codebase is inherit from DETR and ViT of timm. So, 1. ConvertCocoPolysToMask is inherit from DETR, we didn't modify that. 2. This is similar to the original DETR implantation,...