ghimireadarsh / semantic-segmentation-using-unet

As a part of assignment of Machine Vision and Image processing course, a UNET Model has been trained and tested on ISPRS Postdam dataset

Semantic-Segmentation-using-Unet

As a part of assignment of Machine Vision and Image processing course, a UNET Model has been trained and tested on ISPRS Postdam dataset.

Output samples

The complete data visualization, model training, model testing and inference is also available in the notebook.

Best model performance Summary

Total training epochs : 20

Model best performance at 19th epoch

Dataset	Size	Accuracy	Loss
Training set	2000 Images	80.03	0.59
Validation set	200 Images	79.31	0.63
Test set	200 Images	79.99	0.54

Training performance Curve

The model can be further improved by training for more epochs and by changing the hyperparameters.

semantic-segmentation-using-unet's Issues

divide the dataset

Hello, I would like to ask, how do you divide the dataset, I mean training set, validation set, test se