Open JackMa-coder opened 1 year ago
Thanks for your interests. I think the main issue deals with the input images. In the new implementation, the model uses cropped images as inputs (also need to modify camera parameters). I think you could better reproduce AlignSDF using the new repo.
Thanks for your interests. I think the main issue deals with the input images. In the new implementation, the model uses cropped images as inputs (also need to modify camera parameters). I think you could better reproduce AlignSDF using the new repo.
Hello,ruize,i try to use the script in "Dex_YCB dataset organization #8" to generate testset,but i just get 26713 files in meta file path,and 29464 files in other folders,rather than 29466 files. i use test s0 setting. Do you have any idea on the difference in the number of results?
hello, thank you for your great work and patient response. I reproduced the code, the model results related to DexYCB are mostly consistent with the paper, but the Ote indicator is several times (2x ~ 3x) the results in the paper, can you give me some advice? thank you.