Print information about the annotation file. :return:

wrapper for load checkpoint :param prefix: Prefix of model name. :param epoch: Epoch number of model we would like to load. :param co

vertically stack tensors :param tensor_list: list of tensor to be stacked vertically :param pad: label to pad with :return: tensor wi

return a dictionary that contains raw data.

preprocess image and return processed roidb :param roidb: a list of roidb :return: list of img as in mxnet format roidb add new item[

greedily select boxes with high confidence and overlap with current maximum <= thresh rule out overlap >= thresh :param dets: [[x1, y1, x

assign ground truth boxes to anchor positions :param feat_shape: infer output shape :param gt_boxes: assign ground truth :param im_in

Clip boxes to image boundaries. :param boxes: [N, 4* num_classes] :param im_shape: tuple of 2 :return: [N, 4* num_classes]

Get ann ids that satisfy given filter conditions. default skips that filter :param imgIds (int array) : get anns for given imgs

Load anns with the specified ids. :param ids (int array) : integer ids specifying anns :return: anns (object array) : l

turn rpn detection boxes into roidb :param gt_roidb: [image_index]['boxes', 'gt_classes', 'gt_overlaps', 'flipped'] :return:

unmap a subset inds of data into original data of size count

expand from 5 to 4 * num_classes; only the right class has non-zero bbox regression targets :param bbox_targets_data: [k * 5] :param num_

Load model checkpoint from file. :param prefix: Prefix of model name. :param epoch: Epoch number of model we would like to load. :ret

Clip boxes of the pad area. :param tensor: [n, c, H, W] :param pad_shape: [h, w] :return: [n, c, h, w]

Given a vector of widths (ws) and heights (hs) around a center (x_ctr, y_ctr), output a set of anchors (windows).

Return width, height, x center, and y center for an anchor (window).

Convert annotation which can be polygons, uncompressed RLE to RLE. :return: binary mask (numpy 2D array)

append flipped images to an roidb flip boxes coordinates, images will be actually flipped when loading into network :param ro

:param params: dict of str to NDArray :param ctx: the context to convert to :return: dict of str of NDArray with context ctx

remove roidb entries without usable rois

Generate anchor (reference) windows by enumerating aspect ratios X scales wrt a reference (0, 0, 15, 15) window.

Get img ids that satisfy given filter conditions. :param imgIds (int array) : get imgs for given ids :param catIds (int array

prototype for rpn batch: data, im_info, gt_boxes :param roidb: ['image', 'flipped'] + ['gt_boxes', 'boxes', 'gt_classes'] :return: data

Return maximum data and label shape for single gpu

roidb are list, concat them together

merge roidbs into one :param a: roidb to be merged into :param b: roidb to be merged :return: merged imdb

This iter will provide roi and mask data to Mask-RCNN with FPN backbone :param roidb: list of dict, must be preprocessed :par

Remove all boxes with any side smaller than min_size

Prepare ._gts and ._dts for evaluation based on params :return: None

Enumerate a set of anchors for each aspect ratio wrt an anchor.

Enumerate a set of anchors for each scale wrt an anchor.

unmap a subset inds of data into original data of size count

given roidb, add ['assign_level'] :param roidb: roidb to be processed. must have gone through imdb.prepare_roidb

given roidb, add ['bbox_targets'] and normalize bounding box regression targets :param roidb: roidb to be processed. must have gone through i

given roidb, add ['bbox_targets'] and normalize bounding box regression targets :param roidb: roidb to be processed. must have gone through i

given rois, overlaps, gt labels, seg, compute bounding box mask targets :param rois: roidb[i]['boxes'] k * 4 :param overlaps: roidb[i]['m