Can a CNN-based image segmentation algorithm generate variable-length outputs for regression tasks?
I am attempting to develop an image segmentation tool for use with 1D data (identifying different features in data vectors). I have a working example using a UNet architecture, because I like that it is able to handle variable-length inputs without padding or truncating.
Can a CNN-based image segmentation algorithm handle variable-length outputs for regression tasks?
I am attempting to develop an image segmentation tool for use with 1D data (identifying different features in data vectors). I have a working example using a UNet architecture, because I like that it is able to handle variable-length inputs without padding or truncating.