Closed fword closed 7 years ago
@fword No, we did locate that image file in the directory. This is the sha1sum of caption training files: ecafb50a4bd29213772941a5a559d4f741c214b6 Probably you should check if the training file download was completed. Good luck!
This image is also not in my downloaded files and there are only 209961 images in the train set. I modify the script to filter these images in the json file, but not in my downloaded files.
@JinmingZhao maybe your download wasn't completed. I have exactly 210000 images. You should calculate the Sha1 of the zip file you downloaded, and see if it matches with the one provided by the organizers, ecafb50a4bd29213772941a5a559d4f741c214b6
@linoandy yeah, the Sha1 of my downloaded file is not equal to the provided. And I think that ignore these images is a solution to solve this problem if the network is not very stable. Thanks.
NotFoundError: /home/sk/ai_challenger_caption_train_20170902/caption_train_images_20170902/e55ba6db106e61f50802ed3547b325ced2e32a3a.jpg