+- Python Forum (https://python-forum.io)
+-- Forum: Python Coding (https://python-forum.io/forum-7.html)
+--- Forum: Data Science (https://python-forum.io/forum-44.html)
+--- Thread: transfer learning, how preprocess data (/thread-27483.html)
transfer learning, how preprocess data - sveto4ka - Jun-08-2020
Pleae, can you tell me how should I preprosess data while using transfer learning? I mean obligatory preprosessing.
For example, if I want to use resnet50, can I have training images with size 100*100*3? or 512*512*3? Or I should convert them to 224*224*3?
Also, should I normalize data to mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]?
Should I also use /255?
Maybe smth else?
I would be very grateful if you provide explanations and links to articles
RE: transfer learning, how preprocess data - hussainmujtaba - Jun-15-2020
The best way is to use the same preprocessing steps that were used by the authors of resnet50.I guess you should check that paper in detail and you ll have your answer.