python 3.6
pytorch
torchvision
scipy
pillow==6.1
Make data
folder at <this repo's root>/model/
and cd
to this data
folder, then do following tasks:
https://uithcm-my.sharepoint.com/:f:/g/personal/13520797_ms_uit_edu_vn/EolYFvBTBTpOpaAFTd6bw-gBHoMzFSfDnboqZ4-9dDdAOg?e=XDqKfC
curl -L "https://drive.google.com/uc?id=1K12-ySAyNtpJV8f5_3VugeLRfkfKL8iE&export=download" > list.zip
At the data
folder, unzip -q <downloaded zips' files>
(Copy to merge train folders into only 1.)
cp -a add_syn_train/img/. syn_train/img/
cp -a add_syn_train/gt/. syn_train/gt/
cp -a add_syn_train2/img/. syn_train/img/
cp -a add_syn_train2/txt/. syn_train/gt/
cp -a real_train_aug/img/. syn_train/img/
cp -a real_train_aug/txt/. syn_train/gt/
Now, we have the file tree like this:
data/
real/
syn_train/
full_train.txt
real_full.txt
real_train.txt
real_val.txt
syn_train.txt
delete all other directories and zip files.
rm -r syn_train.zip add_syn_train.zip add_syn_train2.zip real_train_aug.zip real.zip list.zip add_syn_train/ add_syn_train2/ real_train_aug/
Go to each model folder, preview the train.py to learn more about the params to fill.
Run python train.py
with params.