Adding Dropout to ResNet50

My goal is to use MC Dropout to ResNet50 architecture. Despite this, it has no dropout layers. What is the correct way to put dropout layers into ResNet50?

1 Like

Did you find an answer to that ?

See here for the PyTorch ResNet implementation:

You would need to define a dropout layer in the init (line 167) and rewrite the _forward_impl(line 266). Then instantiate your custom model and load weights.

it might be true as Resnet 50 not have a dropout but you can easily add a Dropout at after maxpooling in the final fully connected layer. I have added the ref of the timm resnet he has added similarly: