CODEWITHSUNDEEP
kerasneural-network
dimdim dimdim
dimdim dimdim

Reputation: 69

Resnet model parameters

Below i have snippet of python code for Resnet:

base_model = ResNet50(include_top=False, weights='imagenet')

What does include_top=False, weights='imagenet' really mean and is for?

Upvotes: 1

Views: 473

Answers (2)

Dominik Filipiak
Dominik Filipiak

Reputation: 1272

According to the documentation:

  • include_top: whether to include the fully-connected layer at the top of the network.

  • weights: one of None (random initialization), 'imagenet' (pre-training on ImageNet), or the path to the weights file to be loaded.

The keyword include_top was nicely explained in this post. When it comes to weights, setting it to imagenet means using the ResNet50 architecture with ImageNet-trained weights. You can use any other dataset for training and load resulting weights using this argument or just use random ones, though starting with pre-trained ImageNet might be a good idea. Due to their transferability, networks pre-trained on ImageNet are often a good choice to start with.

Upvotes: 1

Simon
Simon

Reputation: 10150

You can find the answer in the docs:

include_top: whether to include the fully-connected layer at the top of the network.

weights: one of None (random initialization), 'imagenet' (pre-training on ImageNet), or the path to the weights file to be loaded.

Upvotes: 1

Related Questions