JeongYeon Nam
JeongYeon Nam
I found perfomance overall enhanced when https://arxiv.org/pdf/2201.11460v1.pdf -> https://arxiv.org/pdf/2201.11460.pdf but checkpoint download link provided is not updated in first commit. checkpoint link seems to be updated to enhcanced version!
https://github.com/long8v/torch_study/tree/master/paper/04_transformer 고쳐주세요
Thanks for great work! I really enjoyed reading your paper. I found that "Transformer" in Table 4 is very strong baseline, and you also provide detailed analysis versus proposed Structured...
Hi there, thank you for providing your work as open source. I can understand your work more thoroughly with your repo. I have a curiosity about HIDDEN_DIM https://github.com/suprosanna/relationformer/blob/scene_graph/configs/scene_2d.yaml#L40 I believe...
Thanks for great work! I am curious about COYO-100M high-quality subset. Are images selected with image resolution? or any other metric such as CLIP score?
> To adapt the corpus to our setting, for each of the 32K test images, we sample a (FOIL, true) pair, and compute the accuracy of each evaluation metric in...
just for your information, example in readme.md returns slightly different values according to PIL version ``` > python clipscore.py example/good_captions.json example/images/ ... CLIPScore: 0.8584 ``` - Pillow == 8.4.0 CLIPScore:...