Splitting data into training and testing dataset in Image segmentation Unet assignment

I wonder why in Unet notebook, we don’t split the data into train and test dataset, and predict mask on test dataset?
Also, isn’t it a data leakage since we predict a mask of input image that is used to train Unet model?


Hi yzhang416,

Good points.

My guess would be that use of a testset is left out of the assignment simply to save on time needed to finish the assignment and the course. The same explanation may hold for using an image from the training set for prediction.

But your points are well-taken. Thanks!

By the way: here is an example in which a test set is used.

1 Like