Hello. I am trying to figure out an issue. I would like to use Resnet architecture in my project. My task is not classification. I am intending to use the Net that takes an image as input and ouput another of same size. Can ResNet50 be used for such task. If yes could someone explained me.
Kind regards
You can use and manipulate the resnet architecture to output a 4-dom tensor. This would be possible by removing the classifier of the model. However, since resnet is downsampling in the feature extraction part of the model you would need to upsample the activations again.
Given than, I would rather just use a segmentation model which already does this.