Given a batch of images (RGB, normalized to [-1...1]) and a specific segmentation class number, returns a tuple with (1) a

Constructs a ResNet-101 model. Args: pretrained (bool): If True, returns a model pre-trained on Places

Constructs a ResNet-101 model. Args: pretrained (bool): If True, returns a model pre-trained on Places

Constructs a ResNet-50 model. Args: pretrained (bool): If True, returns a model pre-trained on Places

Constructs a ResNet-50 model. Args: pretrained (bool): If True, returns a model pre-trained on Places

Constructs a ResNet-101 model. Args: pretrained (bool): If True, returns a model pre-trained on Places

Constructs a ResNet-101 model. Args: pretrained (bool): If True, returns a model pre-trained on Places

Pass a fully-qualified layer name (E.g., module.submodule.conv3) to hook that layer and retain its output each time the model is run.

Runs the dissected model on one batch of the dataset, and returns a multilabel semantic segmentation for the data. Given a ba

Runs the dissected model on one batch of the dataset, and returns a multilabel semantic segmentation for the data. Given a ba

Saves a model by copying tensors to numpy_dict. Converts torch types to numpy types using `t.detach().cpu().numpy()`. Supports saving a s

Utility for scattering semgentations into a one-hot representation.