- Download the raw images and training images from Google Drive
- Extract these folders to the root folder of the project (i.e. you should now have two new folders,
raw
andtraining
) - Run
python make_training.py
to load image segments and classify (NOTE: Images not auto increment! Just make sure you download the most recent from step 1, and upload new versions when done creating new training images) - Run
python train.py
to train the model on lego vs coral vs floor. Note that this only uses 91 images from each class, so you can change. - Run
python3 api.py
to classify segments from an image