Config.py - The hyperparameters of the network customer_init.py - some initialization method Model.py - abstract model readfile.py - read a batch of images from file and send it to tensorflow to train (refer to Tensorflow Dataset) SegnetModel - implementation of segnet model util.py - util methods
-
Make several dirs under the data folder: -- data
-- Logs -- train -- val -- submission
-
download data from kaggle website
-
put the training data under train folder and the test data under val folder
-
You can train your model by writing code in the main function of SegnetModel
if __name__ == "__main__":
segmodel = SegnetModel()
segmodel.training()
EDA: https://colab.research.google.com/drive/12HpbXiewjQdJaZ3qm1QFEBS3xm_lvYbh#scrollTo=cmWKLEOlQSVo training: https://github.com/JasonChu1313/Satellite-Segmentation/tree/master/src