Args: - string

Args: - label_mapping_dict: dictionary from int to int, from original class ID to a new class ID. This is NOT the id_to_class_name dictionary

Duplicate the grayscale channel 3 times. Args: - gray_img: Array with shape (M,N) Returns: - rgb_img: Array with shape (M,N,3)

Generator expression that iterates over the params owned by ``optimizer``. Args: optimizer: An optimizer previously returned from ``

Args: - fpath: string, representing file path

Currently blended mask img background is lime green. Args: - rgb_img: - label_img: - save_fpath - save_to_disk Returns: - Array, r

Given a list of work, and a desired number of n workers, launch n worker processes that will each process 1/nth of the total work. Args: - num

Visualizes a single binary mask by coloring the region inside a binary mask as a specific color, and then blending it with an RGB image. Args:

3x3 convolution with padding

Write out an image explaining the classes inside an image. Args: - label_img - id_to_class_name_map - fname_stem, save_dir Returns: -

Args: - rgb_img: - label_img: - save_fpath - save_to_disk Returns: - Array, representing 3 horizontally concatenated images: from le

Args: - rgb_img: - label_img: - id_to_class_name_map - save_fpath - save_to_disk Returns: - Array, representing 3 horizontally con

Args: - split: string representing training, validation, or testing split of the data - fname_stem: Returns: - img_annot: Python dic

recursively fetches (possible) child module and name of weight to be reparameterized

Args: - x: Numpy array of integers: Returns: - mode of array values (integer)

Quick and dirty method for analysis, not an officially correct implementation

init_hidden()

Call this method to modify your model to take advantage of sparse matrix multiplication. Note that this call alone only augments the model wit

Call this method to monkey patch optimizer step function so that masks can be applied to gradients and weights during training. You mu

Args: - semantic_img: - label_mapping_arr: Returns: - img:

Convert grayscale image to a different grayscale image. Assumption is that there are less than 65,000 classes in the final mapped class for np.uin

Parse a spreadsheet entry from dataset's column, return list of classes in this spreadsheet cell. TODO: handle case for last column in spreadsheet,

Args: - Returns: -

:class:`toRNNBackend`

Function to be called externally for training. Perform fast grayscale->grayscale mapping for training data transformation. Args: - label: P

font_color = (0,0,0) x: x-coordinate from image origin to plot text at y: y-coordinate from image origin to plot text at

Create a Pytorch dataloader from a dataroot and list of relative paths. Args: Returns: - test_loader - data_list: list of 2-tuples (re

Rather than eval unrelabeled model on the univ. relabeled data, we instead map correctness of relabeled model on relabeled univ. data

take in array of IoU/Acc., return non-excluded IoU/acc values

Rather than feeding in crops w/ sliding window across the full-res image, we downsample/upsample the image to a default inference size. This may

Load 2-tuple of image data from disk (RGB and label). Args: - fname_stem: string representing Returns: - rgb_img - label_img

Encoding described here: https://github.com/cocodataset/panopticapi/blob/master/panopticapi/utils.py#L30 "Given semantic category uniq

Since the class masks to instance masks don't match up exactly, and are provided in images with very different resolutions, we have to take the m

Args: - None Returns: - iou_class: Array - accuracy_class: Array - m

Given a grayscale image where intensities denote instance IDs (same intensity denotes belonging to same instance), convert this to an RGB image whe

It is expected that all threads in the CTA enter this function!

Compute IoU on Numpy arrays on CPU. We will be reasoning about each matrix cell individually, so we can reshape (flatten) these arrays in

Must return the number of data points

Args: - dataset_name: str Returns: - Numpy array of shape (N,3) with RGB tuple corresponding to each class.

TODO: may need to make a copy here, if it won't make one for us. Args: - semantic_img: Pytorch CPU long tensor representing (M,N) matrix,

Cannot be a blank string. Args: - uentry: string, representing TSV entry from `Universal` taxonomy column Returns: - full_name: exact dupli

Use 1x1 convolution to create linear mapping P of each pixel's probabilities to a new space. The matrix weights Pij are binary 0/1 values and ar

Args: - txt_fpath: string representing path to txt file Returns: - txt_lines: list of strings, one per line of file

reset_parameters()

Args: - image: Numpy array of shape () - scale: Returns: - image_scaled:

7-tuple consists of (1-3) rgb mask 3-sequence for label, (4-6) rgb mask 3-sequence for predictions, (7) color palette Args: - img_rg

* Returns the number of entries in the database * @return number of entries in the database */

Performs a single optimization step. Arguments: closure (callable, optional): A closure that reevaluates the model

Args: - label_img: label map before any update has taken place. - segment_mask: 0/1 binary image showing segment pixels - old_val: old pixel

Args: - Returns: -

Depending on the brightness_factor, gives a lighter or darker color i.e. a color with less or more saturation than the original color