Apply pytorch's to() to all tensors in this container.

Load instance masks for the given image. Different datasets use different ways to store masks. Override this method to load instance

Apply the given mask to the image.

Load the specified image and return a [H,W,3] Numpy array.

Computes IoU overlaps between two sets of boxes. boxes1, boxes2: [N, (y1, x1, y2, x2)].

Load configurations in config_fn as class attributes of this class. Args: config_fn: YAML file containing configuration.

Apply pytorch's cpu() to all tensors in this container.

Train the model. train_dataset, val_dataset: Training and validation Dataset objects. learning_rate: The learning rate to train with

Modified version of the correspoding Keras function with the addition of multi-GPU support and the ability to exclude some layers from loading

Apply pytorch's numpy() to all tensors in this container.

Call in a loop to create terminal progress bar. Args: losses: Loss object

Implements ROI Pooling on multiple levels of the feature pyramid. Params: - pool_size: [height, width] of the output pooled regions. Usually

Generate random colors. To get visually distinct colors, generate them in HSV space then convert to RGB.

Resizes an image keeping the aspect ratio unchanged. min_dim: if provided, resizes the image such that it's smaller dimension == min_dim

Apply same indexing to all tensors in container

Sets the model log directory and epoch counter. model_path: If None, or a format different from what this code uses then set a new log di

Reformats the detections of one image from the format of the neural network output to a format suitable for use in the rest of the application

Convert a dictionary to class attributes in a tree-fashion. The root object is Config class. Note: Recursive function. Args:

Compute Intersection over Union of ground truth and predicted masks. gt_masks and pred_masks must be in mini-mask format. Boxes indicate mask

Compute mean average precision.

Compute the intersections between box1 and box2 and return the indexes of the intersection box relative to both boxes. Note: Boxes co

This function returns a printer function, that prints a message and then print a tensor. Used by register_hook in the backward pass.

For each ROI, takes TOP probs, ids and deltas.

Trains a single epoch.

Validation step. Usually called with torch.no_grad().

Stores min, max and avg box sides (for height and width)

Build Mask R-CNN architecture.

Given the anchors and GT boxes, compute overlaps and identify positive anchors and deltas to refine them to match their corresponding GT boxes.

All configuration checks must be placed here.

boxes: [N, 4] each col is y1, x1, y2, x2 window: [4] in the form y1, x1, y2, x2

Calculates IoU of the given box with the array of the given boxes. box: 1D vector [y1, x1, y2, x2] boxes: [boxes_count, (y1, x1, y2, x2)]

Compute Intersection over Union of ground truth and predicted masks. Args: gt_masks (torch.IntTensor((img_height, img_width, nb_gt_masks)

Compute mean average precision. Args: ious (torch.FloatTensor((nb_gt_masks, nb_pred_masks))): Intersection over Union. R

Loss for Mask R-CNN bounding box refinement. target_bbox: [batch, num_rois, (dy, dx, log(dh), log(dw))] target_class_ids: [batch, num_rois].

Loss for the classifier head of Mask RCNN. target_class_ids: [batch, num_rois]. Integer class IDs. Uses zero padding to fill in the array

Mask binary cross-entropy loss for the masks head. target_masks: [batch, num_rois, height, width]. A float32 tensor of values 0 or 1. Use

Return the RPN bounding box loss graph. target_bbox: [batch, max positive anchors, (dy, dx, log(dh), log(dw))]. Uses 0 padding to fill in

RPN anchor classifier loss. rpn_match: [batch, anchors, 1]. Anchor match type. 1=positive, -1=negative, 0=neutral anchor. rpn_

Runs the detection pipeline. images: Image Returns a list of dicts, one dict per image. The dict contains: rois: [N, (y1, x1

Refine classified proposals and filter overlaps and return final detections. Inputs: rois: [N, (y1, x1, y2, x2)] in normalized coordi

Subsamples proposals and generates target box refinement, class_ids, and masks for each. Inputs: proposals: [batch, N, (y1, x1, y2, x2)]

Display the given set of images, optionally with titles. images: list or array of image tensors in HWC format. titles: optional. A list of tit

boxes: [num_instance, (y1, x1, y2, x2, class_id)] in image coordinates. masks: [height, width, num_instances] class_ids: [num_instances]

Display values in a table format. table: an iterable of rows, and each row is an iterable of values.

Filter results according to stats.

Finds the last checkpoint file of the last trained model in the model directory. Returns: log_dir: The directory where events and weig

Blocks the configuration so it cannot be modified. Used to prevent changes in the configuration during execution.

scale: 1D array of anchor sizes in pixels. Example: [32, 64, 128] ratios: 1D array of anchor ratios of width/height. Example: [0.5, 1, 2]

Generate anchors at different levels of a feature pyramid. Each scale is associated with a level of the pyramid, but each ratio is used in all

Initialize model weights.

Load configurations in config_fn as class attributes of this class. Args: config_fn: PATH to YAML file containing configu

Load and return ground truth data for an image (image, mask, bounding boxes). augment: If true, apply random image augmentation. Currently

Load a subset of the nuclei dataset. dataset_dir: Root directory of the dataset subset: Subset to load. Either the name of the sub-di

Encodes instance masks to submission format.

Merge configuration present at config_fn into this class. Args: config_fn: YAML file containing configuration.

Takes RGB images with 0-255 values and subtraces the mean pixel and converts it to float. Expects image colors in RGB order.

Prepares the Dataset class for use. TODO: class map is not supported yet. When done, it should handle mapping classes from diff

Receives anchor scores and selects a subset to pass as proposals to the second stage. Filtering is done based on anchor scores and non-max sup

Encodes a mask in Run Length Encoding (RLE). Returns a string of space-separated values.