seq_id is only provided so that all other datasets can share a common API.

Args: - segmentid: integer representing segment unique ID - fname_stem - split: dataset split, i.e. 'train' or 'val' Returns: - se

Get COCO Panoptic semantic annotations from .json file Args: - split: string representing training, validation, or testing split of the data

1. Take the reciprocal of all numbers in the dataset 2. Find the arithmetic mean of those reciprocals 3. Take the reciprocal of that number

Horizontally stack a left image with a palette image on the right.

Returns: - RGB 3 channel nd-array with shape H * W * 3

init_inference()

For training, we use a model pretrained on ImageNet. Irrelevant at inference. Args: - pretrained_fpath: str representing pa

Returns the L2 distance between the vector "p1[0:size-1]" and the data point with index "idx_p2" stored in the class:

* Find the "num_closest" nearest neighbors to the \a query_point[0:dim-1]. Their indices are stored inside * the result object. * \sa radiusSea

Args: - label_img_rgb: Returns: - label_img

Load Pytorch pre-trained model from disk of type torch.nn.DataParallel. Note that `args.num_model_classes` will be size of logits output.

Loads a state_dict created by an earlier call to state_dict(). If ``fp16_optimizer_instance`` was constructed from some ``init_optim

We use the MSeg dataroot explicitly, as specified in mseg/utils/dataset_config.py

Given the full tsv with all dataset->universal mappings, and given relabeled update records, we form new "relabeled taxonomies" for each dataset, a

We use the MSeg dataroot explicitly, as specified in mseg/utils/dataset_config.py

Convert a json file path e.g. gtFine/train/1/210316_gtFine_polygons.json to: gtFine/train/1/210316_gtFine_labelids.png

Visualize masks of a chosen category from BDD. Usage: python BDDImageLevelDataset.py \ --dataroot /Users/johnlamb/Downloads/bdd100k --classname

Visualize masks of a chosen category from SUN RGB-D. Usage: python SunrgbdImageLevelDataset.py \ --dataroot /Users/johnlamb/Downloads/SUNRGBD-37

Example Use: python MapillaryMaskDataset.py -classname="Ground Animal" --folder_prefix mseg_mapillary_water_2019_04_19 --num_processes 4 --dataroo

Visualize masks of a chosen category from COCO Panoptic. Interesting choices: 'floor','curtain','mountain','rug','hill' 'light', 'lamp', 'chan

Visualize masks of a chosen category from COCO Panoptic. Interesting choices: 'cabinet-merged', 'rug-merged', 'road', 'bus', 'truck', 'cup','

Args: - split: string representing split of data set to use, must be either 'train','val','test' - data_root: path to whe

* Returns a pointer to a piece of new memory of the given size in bytes * allocated from the pool. */

Args: - segment_mask - label_img Returns: - is_single_class - sem_class_ids

Returns the stringified extra_repr() of a module.

Feed input through the network. In addition to running a crop through the network, we can flip the crop horizontally, run both crops through t

Returns: (H, W) array, a mask for all pixels that have a prediction

Pass in by reference Torch tensor, and normalize its values. Args: - input: Torch tensor of shape (3,M,N), must be in this order,

Print usage and parse arguments.

Parse an entry from dataset element, return list of classes. Args: - entry: string, representing cell in taxonomy spreadsheet Returns: - cl

Can be a blank string, e.g. ''. We ensure no additional whitespace present. Args: - test_entry Returns: - test_entry

Args: - uentry: string, representing TSV entry from `Universal` taxonomy column Returns: - level: integer, representing depth of node in tre