A single linear layer which does a 1x1 conv

Rescales the input PIL.Image to the given 'size'. 'size' will be the size of the smaller edge. For example, if height > width, then image wil

Converts a PIL.Image (RGB) or numpy.ndarray (H x W x C) in the range [0, 255] to a torch.FloatTensor of shape (C x H x W) in the range [0.0, 1.0]

r""" Adversarial loss https://arxiv.org/abs/1711.10337

takes 2 distances, puts through fc layers, spits out value between [0,1] (if use_sigmoid is True)

Randomly horizontally flips the given PIL.Image with a probability of 0.5

Random crop the given PIL.Image to a random size of (0.08 to 1.0) of the original size and and a random aspect ratio of 3/4 to 4/3 of the original

Inception-v1 I3D architecture. The model is introduced in: Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset Joa

Training process logger Note: Used by BaseTrainer to save training history.

Trainer class Note: Inherited from BaseTrainer.

Base class for all data loaders

Base class for all models

Base class for all trainers

Use these inner classes to mimic 3D operation by using 2D operation frame by frame.

Pre-trained network with all channels equally weighted by default