scale x to the same size as y

Custom Norm Function to allow flexible switching

Initialize Model Weights

Wrapper Around the Upsample Call

Retrieve the network trunk and channel counts.

Create aspp block

scale x by some factor

Set immutability to is_immutable and recursively apply the setting to all nested AttrDicts.

Wrapper Around the Upsample Call

For class uniform sampling ... every epoch, we want to recompute which tiles from which images we want to sample from, so that the

format scale name :prefix: a string that is the beginning of the field name :scale: a scale value (0.25, 0.5, 1.0, 2.0)

Parameters ---------- block (nn.Module): Bottleneck class. - For SENet154: SEBottleneck - For SE-ResN

:img_label_pairs: A list of pairs of [img,label]

Handle partial loading when some tensors don't match up in size. Because we want to use models that were trained off a different number o

Run validation for one epoch :val_loader: data loader for validation :net: the network :criterion: loss fn :optimizer: optimizer

Hierarchical form of attention where we only predict attention for pairs of scales at a time. At inference time we can combi

Run the network, and return final feature and logit predictions

Run the network, and return final feature and logit predictions

Add margin around an image top, right, bottom, left are the margin widths, in pixels margin_color is what to use for the margins

Calculate weights of classes based on the training crop

Calculate weights of the classes based on training crop

3x3 convolution with padding

3x3 convolution with padding

Custom Softmax

dump a single batch of images :dump_dict: a dictionary containing elements to dump out 'input_images': source image

Modified IOU mechanism for on-the-fly IOU calculations ( prevents memory overflow for large dataset) Only applies to eval/eval.py

Find image and segmentation mask files and return a list of tuples of them. Inputs: img_root: path to parent directo

Flip Tensor along a dimension

Using the sigmiod operation both. Args: logits_4D : [N, C, H, W], dtype=float32 labels_4D : [N, H,

Initialize Pretrain Model Information, Dowload weights, load weights, set variables

Create Filename list for images in the provided path input: path to directory with *only* images files returns: items list with None fil

Randomly sample num items from the list alist: list of centroids to sample from num: can be larger than the list and if so, then wrap aro

From a segmentation, compute a binary boundary map with 1 pixel wide boundaries. The boundary pixels are offset by 1/2 pixel towards the origin f

Combine multiple scales of predictions together with attention predicted jointly off of multi-scale features.

Adjust brightness of an Image. Args: img (PIL Image): PIL Image to be adjusted. brightness_factor (float): How much to adjust th

Adjust contrast of an Image. Args: img (PIL Image): PIL Image to be adjusted. contrast_factor (float): How much to adjust the con

Adjust hue of an image. The image hue is adjusted by converting the image to HSV and cyclically shifting the intensities in the hue channel (