Regression vs Classification

MostefaBen / Fully-automatic-brain-tumor-segmentation-with-deep-learning-based-selective-attention

Keras implementation of the paper "Fully automatic brain tumor segmentation with deep learning-based selective attention using overlapping patches and multi-class weighted cross-entropy", by Mostefa Ben naceur et al. (Medical Image Analysis)

MIT License

19 stars 4 forks source link

Regression vs Classification #2

Open xav69ier opened 3 years ago

xav69ier commented 3 years ago

I have my DataSet of co-registered images X=(240,240,4) and Y=(240,240,1) I will try out your model, but first I need to build an array of overlapping patches - but I think I should also consider to change your loss (categorical_crossentropy) to (mean_square_error) and my activation at the output of the adapted model should be (relu) instead of (softmax) - considering that my problem is regression and not classification -

xav69ier commented 3 years ago

... and of course I will prepare an array of patches departing from the array of image, but I wonder in which order I should arrange the patches that make together a whole slice - and also I will need to stack patches corresponding to 4 channels (T1C,T1,T2,FLAIR)