Holds the configuration for anything you want it to. To get the currently active config, call get_cfg(). To use, just do cfg.x instead o

Composes several augmentations together. Args: transforms (List[Transform]): list of transforms to compose. Example: >>> augme

Transform that does all operations on the GPU for super speed. This doesn't suppport a lot of config settings and should only be used for pro

Keeps an average window of the specified number of items.

Stores all the information necessary to calculate the AP for one IoU and one class. Note: I type annotated this because why not.

Transforms a BRG image made of floats in the range [0, 255] to whatever input the current backbone network needs. transform is a transfo

Transorm to be used when evaluating.

If preserve_aspect_ratio is true, this resizes to an approximate area of max_size * max_size

Transform to be used when training.

Why is this a class? Why do I have a class for creating and parsing save paths? What am I doing with my life? --- said D. Bolya

This is a custom version of DataParallel that works better with our training data. It should also be faster than the general case.

Depth gradient Loss for instance segmentation

A wrapper for running the network and computing the loss This is so we can more efficiently use DataParallel.

Joint Weighted Loss Function for PlaneRecNet Compute Targets: 1) Classification loss 2) Segmentation loss 3) Depth estima

A simple progress bar that just outputs a string.

Adapted from torchvision.models.resnet

Pads the image to the input width and height, filling the background with mean and putting the image in the top-left. Note: this expects

The general class for reading training and validation datasets. Data sample is organized with a extend format of COCO dataset. https://coco

Adapted from torchvision.models.resnet

Resize the image to its long side == cfg.max_size, filling the area: [(long side - short side) * long side] to with mean and putting the

Transforms a tensorized image by swapping the channels in the order specified in the swap tuple. Args: swaps (int triple): final orde

The original Virtual Normal Loss Function with some modification. Since we can't assume that every image are taken by the same camera.

A class that lets you go: with timer.env(fn_name): # (...) That automatically manages a timer start and stop for you.