File size: 8,174 Bytes
506da10 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 |
# coding=utf-8
# Copyright 2021 The Deeplab2 Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Provides data from segmentation datasets.
Currently, we support the following datasets:
1. Cityscapes dataset (https://www.cityscapes-dataset.com).
The Cityscapes dataset contains 19 semantic labels (such as road, person, car,
and so on) for urban street scenes.
2. KITTI-STEP (http://www.cvlibs.net/datasets/kitti/).
The KITTI-STEP enriches the KITTI-MOTS data with additional `stuff'
anntotations.
3. MOTChallenge-STEP (https://motchallenge.net/).
The MOTChallenge-STEP enriches the MOTSChallenge data with additional `stuff'
annotations.
4. MSCOCO panoptic segmentation (http://cocodataset.org/#panoptic-2018).
Panoptic segmentation annotations for MSCOCO dataset. Note that we convert the
provided MSCOCO panoptic segmentation format to the following one:
panoptic label = semantic label * 256 + instance id.
5. Cityscapes-DVPS (https://github.com/joe-siyuan-qiao/ViP-DeepLab)
The Cityscapes-DVPS dataset augments Cityscapes-VPS
(https://github.com/mcahny/vps) with depth annotations.
References:
- Marius Cordts, Mohamed Omran, Sebastian Ramos, Timo Rehfeld, Markus
Enzweiler, Rodrigo Benenson, Uwe Franke, Stefan Roth, and Bernt Schiele, "The
Cityscapes Dataset for Semantic Urban Scene Understanding." In CVPR, 2016.
- Andreas Geiger and Philip Lenz and Raquel Urtasun, "Are we ready for
Autonomous Driving? The KITTI Vision Benchmark Suite." In CVPR, 2012.
- Alexander Kirillov, Kaiming He, Ross Girshick, Carsten Rother, and Piotr
Dollar, "Panoptic Segmentation." In CVPR, 2019.
- Tsung-Yi Lin, Michael Maire, Serge J. Belongie, Lubomir D. Bourdev, Ross B.
Girshick, James Hays, Pietro Perona, Deva Ramanan, Piotr Dollar, and C.
Lawrence Zitnick, "Microsoft COCO: common objects in context." In ECCV, 2014.
- Anton Milan, Laura Leal-Taixe, Ian Reid, Stefan Roth, and Konrad Schindler,
"Mot16: A benchmark for multi-object tracking." arXiv:1603.00831, 2016.
- Paul Voigtlaender, Michael Krause, Aljosa Osep, Jonathon Luiten, Berin
Balachandar Gnana Sekar, Andreas Geiger, and Bastian Leibe. "MOTS:
Multi-object tracking and segmentation." In CVPR, 2019
- Mark Weber, Jun Xie, Maxwell Collins, Yukun Zhu, Paul Voigtlaender, Hartwig
Adam, Bradley Green, Andreas Geiger, Bastian Leibe, Daniel Cremers, Aljosa
Osep, Laura Leal-Taixe, and Liang-Chieh Chen, "STEP: Segmenting and Tracking
Every Pixel." arXiv: 2102.11859, 2021.
- Dahun Kim, Sanghyun Woo, Joon-Young Lee, and In So Kweon. "Video panoptic
segmentation." In CVPR, 2020.
- Siyuan Qiao, Yukun Zhu, Hartwig Adam, Alan Yuille, and Liang-Chieh Chen.
"ViP-DeepLab: Learning Visual Perception with Depth-aware Video Panoptic
Segmentation." In CVPR, 2021.
"""
import collections
# Dataset names.
_CITYSCAPES = 'cityscapes'
_CITYSCAPES_PANOPTIC = 'cityscapes_panoptic'
_KITTI_STEP = 'kitti_step'
_MOTCHALLENGE_STEP = 'motchallenge_step'
_CITYSCAPES_DVPS = 'cityscapes_dvps'
_COCO_PANOPTIC = 'coco_panoptic'
# Colormap names.
_CITYSCAPES_COLORMAP = 'cityscapes'
_MOTCHALLENGE_COLORMAP = 'motchallenge'
_COCO_COLORMAP = 'coco'
# Named tuple to describe dataset properties.
DatasetDescriptor = collections.namedtuple(
'DatasetDescriptor', [
'dataset_name', # Dataset name.
'splits_to_sizes', # Splits of the dataset into training, val and test.
'num_classes', # Number of semantic classes.
'ignore_label', # Ignore label value used for semantic segmentation.
# Fields below are used for panoptic segmentation and will be None for
# Semantic segmentation datasets.
# Label divisor only used in panoptic segmentation annotation to infer
# semantic label and instance id.
'panoptic_label_divisor',
# A tuple of classes that contains instance annotations. For example,
# 'person' class has instance annotations while 'sky' does not.
'class_has_instances_list',
# A flag indicating whether the dataset is a video dataset that contains
# sequence IDs and frame IDs.
'is_video_dataset',
# A string specifying the colormap that should be used for
# visualization. E.g. 'cityscapes'.
'colormap',
# A flag indicating whether the dataset contains depth annotation.
'is_depth_dataset',
]
)
CITYSCAPES_INFORMATION = DatasetDescriptor(
dataset_name=_CITYSCAPES,
splits_to_sizes={'train_fine': 2975,
'train_coarse': 22973,
'trainval_fine': 3475,
'trainval_coarse': 23473,
'val_fine': 500,
'test_fine': 1525},
num_classes=19,
ignore_label=255,
panoptic_label_divisor=None,
class_has_instances_list=None,
is_video_dataset=False,
colormap=_CITYSCAPES_COLORMAP,
is_depth_dataset=False,
)
CITYSCAPES_PANOPTIC_INFORMATION = DatasetDescriptor(
dataset_name=_CITYSCAPES_PANOPTIC,
splits_to_sizes={'train_fine': 2975,
'val_fine': 500,
'trainval_fine': 3475,
'test_fine': 1525},
num_classes=19,
ignore_label=255,
panoptic_label_divisor=1000,
class_has_instances_list=tuple(range(11, 19)),
is_video_dataset=False,
colormap=_CITYSCAPES_COLORMAP,
is_depth_dataset=False,
)
KITTI_STEP_INFORMATION = DatasetDescriptor(
dataset_name=_KITTI_STEP,
splits_to_sizes={'train': 5027,
'val': 2981,
'test': 11095},
num_classes=19,
ignore_label=255,
panoptic_label_divisor=1000,
class_has_instances_list=(11, 13),
is_video_dataset=True,
colormap=_CITYSCAPES_COLORMAP,
is_depth_dataset=False,
)
MOTCHALLENGE_STEP_INFORMATION = DatasetDescriptor(
dataset_name=_MOTCHALLENGE_STEP,
splits_to_sizes={'train': 525, # Sequence 9.
'val': 600, # Sequence 2.
'test': 0},
num_classes=7,
ignore_label=255,
panoptic_label_divisor=1000,
class_has_instances_list=(4,),
is_video_dataset=True,
colormap=_MOTCHALLENGE_COLORMAP,
is_depth_dataset=False,
)
CITYSCAPES_DVPS_INFORMATION = DatasetDescriptor(
dataset_name=_CITYSCAPES_DVPS,
# The numbers of images are 2400/300/300 for train/val/test. Here, the
# sizes are the number of consecutive frame pairs. As each sequence has 6
# frames, the number of pairs for the train split is 2400 / 6 * 5 = 2000.
# Similarly, we get 250 pairs for the val split and the test split.
splits_to_sizes={'train': 2000,
'val': 250,
'test': 250},
num_classes=19,
ignore_label=255,
panoptic_label_divisor=1000,
class_has_instances_list=tuple(range(11, 19)),
is_video_dataset=True,
colormap=_CITYSCAPES_COLORMAP,
is_depth_dataset=True,
)
COCO_PANOPTIC_INFORMATION = DatasetDescriptor(
dataset_name=_COCO_PANOPTIC,
splits_to_sizes={'train': 118287,
'val': 5000,
'test': 40670},
num_classes=134,
ignore_label=0,
panoptic_label_divisor=256,
class_has_instances_list=tuple(range(1, 81)),
is_video_dataset=False,
colormap=_COCO_COLORMAP,
is_depth_dataset=False,
)
MAP_NAME_TO_DATASET_INFO = {
_CITYSCAPES: CITYSCAPES_INFORMATION,
_CITYSCAPES_PANOPTIC: CITYSCAPES_PANOPTIC_INFORMATION,
_KITTI_STEP: KITTI_STEP_INFORMATION,
_MOTCHALLENGE_STEP: MOTCHALLENGE_STEP_INFORMATION,
_CITYSCAPES_DVPS: CITYSCAPES_DVPS_INFORMATION,
_COCO_PANOPTIC: COCO_PANOPTIC_INFORMATION,
}
MAP_NAMES = list(MAP_NAME_TO_DATASET_INFO.keys())
|