sergevkim / facesclassification

FacesClassification

Classifier for CelebaHQ dataset

1) Download and prepare data

• !wget <- TODO

• python scripts/prepare_labels.py <- TODO put labels.txt in zip file

2) Run train

• python scripts/run_train.py

OR

• run jupyter notebook and run all cells in train.ipynb

facesclassification's Issues

class Trainer breaks training

in 2-draft.ipynb notebook the first three cells don't fit the model, but the last cell contains the program that fits it perfectly. I see no difference between them, however, it exists.

use PIL

cv2 -> PIL

ResNet classifiers

forward_hook, setattr