proto*/

proto*/

Compute bounding-box regression targets for an image. Parameters: ----------- ex_rois: ROIs from external source (selective search or

Detect object classes in an image given object proposals. Arguments: im (ndarray): color image to test (in BGR order) Ret

Return the default path where PASCAL VOC is expected to be installed.

Builds an input blob from the images in the roidb at the specified scales.

Builds an input blob from the images in the roidb at the specified scales.

Return the blobs to be used for the next minibatch.

Return the blobs to be used for the next minibatch.

Load the indexes listed in this dataset's image set file. Examples path is: self._devkit_path + /VOCdevkit2007/VOC2007/ImageS

Load image and bounding boxes info from XML file in the PASCAL VOC format according to image index

Load gt_masks information from SBD's additional data

Merge user's config into default config dictionary, clobbering the options in b whenever they are also specified in a. Need to ensure

Returns: list, each list is a dictionary contains mask list, box list

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

Generate a random sample of RoIs comprising foreground and background examples.

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

Write results as a pkl file, note this is different from detection task since it's difficult to write masks to txt

Add information needed to train bounding-box regressors.

This method is only accessed when we use maskdb, so implement here Append flipped images to mask database Note this method do

This method is irrelevant with database, so implement here Append flipped images to ROI database Note this method doesn't act

Apply non-maximum suppression to all predicted boxes output by the test_net method.

only implement single maskdb now

only implement single roidb now

Args: cache_dir: output directory for cached mask annotation devkit_path: root directory of VOCdevkitSDS image_names: use

Compute bbox mean and stds for mcg proposals Since mcg proposal are stored on disk, so we precomputed it here once and save them to disk

Wrapper function for mask voting, note we already know the class of boxes and masks Args: masks: ~ n x mask_sz x mask_sz boxe

inject deep into distutils to customize how the dispatch to gcc/nvcc works. If you subclass UnixCCompiler, it's not trivial to get your subcl

Find a file in a search path

During forward, we need to do several things: 1. Apply bounding box regression output which has highest classification sco

Locate the CUDA environment on the system Returns a dict with keys 'home', 'nvcc', 'include', and 'lib64' and values giving the absolute path

This function implements mask voting mechanism to give finer mask n is the candidate boxes (masks) number Args: masks: All masks

Parse input arguments

Parse input arguments

Parse input arguments

Parse input arguments.

Get cooresponding masks, boxes, classes according to image name Args: image_name: input image name devkit_path: root dir for

Parse a PASCAL VOC xml file

Convert rois to network input support multi-scale testing

Enrich the imdb's roidb by adding some derived quantities that are useful for training. This function pre-computes the maximum overla

Set the roidb to be used by this layer during training.

This function plot segmentation results to specific directory Args: img_names: list