cnn fast ai using superman bird data set I downloaded from google
General steps
- Download superman and bird images from google using someone's python scripts
- Delete bad images
- Start formatting the filenames using a simple python script I wrote
- Separate the files into training and validation folders
- upload to this github
- clone back to my google cloud
- run the jupyter notebook
- experiment with batch size since the images are not a lot
- also experiment with data augmentation, and the settings for the learn including annealing and differential learning rate