Pytorch CNN models for recognition of photographic style
python3 predict_local.py -i DOF5.jpg
results in
Color
Colorful 0.7939348220825195
Composition
Undefined 0.40584397315979004
Rule of Thirds 0.2752402722835541
Centered 0.21816390752792358
DoF
Shallow 0.7070949077606201
Palette
Gray 0.4880247116088867
Blue 0.43242210149765015
White 0.3236027657985687
Brown 0.1656956821680069
Green 0.16228264570236206
Type
Landscape 0.22733041644096375
Cityscape 0.10748865455389023
predict_local.py
downloads the model the first time it is called, but you can also find the pretrained models here:
https://drive.google.com/drive/folders/13vd2JQ6wsvKuTovoOc_TjWwzCdDcIHIi?usp=sharing