I wanted to use the resnet101 for a regression like problem. So, the input of the network is a image with an associated target (a number), and I want to get an output by training a model like regression.
What I am doing is adding a linear layer in the end of the resnet101 so the output if a single value.
model = models.resnet101(pretrained=True)
num_ftrs = model.fc.in_features
model.fc = nn.Linear(num_ftrs, 1)
Does this make sense? How would you suggest to use resnet or other pre trained cnn for continuous data?