Preparing images data - Githubissues

menyifang / ADGAN

The Implementation of paper "Controllable Person Image Synthesis with Attribute-Decomposed GAN" CVPR 2020 (Oral); Pose and Appearance Attributes Transfer;

473 stars 90 forks source link

Preparing images data #28

Open Ridam2k opened 3 years ago

Ridam2k commented 3 years ago

Hi I'm having trouble preparing the images downloaded from DeepFashion so that they match the entries in train.lst and test.lst. The downloaded data contains images organised in a number of sub-folders, and even on flattening the directory the image names don't match the train.lst and test.lst values. Also, the google drive link provided for generated images contains only testing images. Any help regarding how to proceed further would be greatly appreciated!

lining-7 commented 3 years ago

Have you solved this problem? I had the same problem。

sandammeer commented 3 years ago

Ive got the same problem

confifu commented 3 years ago

You need to get rid of the _ after id and the last _ . Assuming parent directory is named fashion as mentioned earlier, in ubuntu I did this:

find fashion -type f -exec sh -c 'new=$(echo "{}" | tr -d "/"|sed s/id_/id/); mv "{}" "$new"' \; : To flatten the directory fashion and get rid of _ after id.
for f in *.jpg; do mv "$f" "$(echo "$f" | sed 's/$.*$_/\1/')"; done : To get rid of the last _