According to the published paper, the ImageNet-Compatible dataset is used to conduct experiments. By downloading the above dataset from the website in reference [1], it is strange to find that there are only 904 images can be downloaded, rather than 1000 images illustrated in the paper. Therefore, could you please validate how many images were actually used in the experimental phase? Of course, if all 1000 images were used in your experiments, could you please share these images through Google Clouds or Baidu Clouds (urls of the missed 96 images are invalid now)?
According to the published paper, the ImageNet-Compatible dataset is used to conduct experiments. By downloading the above dataset from the website in reference [1], it is strange to find that there are only 904 images can be downloaded, rather than 1000 images illustrated in the paper. Therefore, could you please validate how many images were actually used in the experimental phase? Of course, if all 1000 images were used in your experiments, could you please share these images through Google Clouds or Baidu Clouds (urls of the missed 96 images are invalid now)?
Thanks very much!