SatwikKambham commited on
Commit
ff7f0d1
1 Parent(s): 43f849e

Added loading script and updated readme

Browse files
Files changed (2) hide show
  1. README.md +16 -0
  2. suim.py +132 -0
README.md CHANGED
@@ -1,3 +1,19 @@
1
  ---
2
  license: mit
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  ---
 
1
  ---
2
  license: mit
3
+ dataset_info:
4
+ config_name: suim
5
+ features:
6
+ - name: img
7
+ dtype: image
8
+ - name: mask
9
+ dtype: image
10
+ splits:
11
+ - name: train
12
+ num_bytes: 511917
13
+ num_examples: 1525
14
+ - name: test
15
+ num_bytes: 35774
16
+ num_examples: 110
17
+ download_size: 183261195
18
+ dataset_size: 547691
19
  ---
suim.py ADDED
@@ -0,0 +1,132 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Copyright 2020 The HuggingFace Datasets Authors and the current dataset
2
+ # script contributor.
3
+ #
4
+ # Licensed under the Apache License, Version 2.0 (the "License");
5
+ # you may not use this file except in compliance with the License.
6
+ # You may obtain a copy of the License at
7
+ #
8
+ # http://www.apache.org/licenses/LICENSE-2.0
9
+ #
10
+ # Unless required by applicable law or agreed to in writing, software
11
+ # distributed under the License is distributed on an "AS IS" BASIS,
12
+ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13
+ # See the License for the specific language governing permissions and
14
+ # limitations under the License.
15
+ """Semantic Segmentation of Underwater IMagery (SUIM) dataset"""
16
+
17
+
18
+ import os
19
+
20
+ import datasets
21
+
22
+
23
+ _CITATION = """\
24
+ @inproceedings{islam2020suim,
25
+ title={{Semantic Segmentation of Underwater Imagery: Dataset and Benchmark}},
26
+ author={Islam, Md Jahidul and Edge, Chelsey and Xiao, Yuyang and Luo, Peigen and Mehtaz,
27
+ Muntaqim and Morse, Christopher and Enan, Sadman Sakib and Sattar, Junaed},
28
+ booktitle={IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
29
+ year={2020},
30
+ organization={IEEE/RSJ}
31
+ }
32
+ """
33
+
34
+ _DESCRIPTION = """\
35
+ The SUIM dataset is a dataset for semantic segmentation of underwater imagery.
36
+
37
+ The dataset consists of 1525 annotated images for training/validation and
38
+ 110 samples for testing.
39
+
40
+ | Object category | Symbol | RGB color code |
41
+ |----------------------------------|--------|----------------|
42
+ | Background (waterbody) | BW | 000 (black) |
43
+ | Human divers | HD | 001 (blue) |
44
+ | Aquatic plants and sea-grass | PF | 010 (green) |
45
+ | Wrecks and ruins | WR | 011 (sky) |
46
+ | Robots (AUVs/ROVs/instruments) | RO | 100 (red) |
47
+ | Reefs and invertebrates | RI | 101 (pink) |
48
+ | Fish and vertebrates | FV | 110 (yellow) |
49
+ | Sea-floor and rocks | SR | 111 (white) |
50
+
51
+
52
+ For more information about the original SUIM dataset,
53
+ please visit the official dataset page:
54
+
55
+ https://irvlab.cs.umn.edu/resources/suim-dataset
56
+
57
+ Please refer to the original dataset source for any additional details,
58
+ citations, or specific usage guidelines provided by the dataset creators.
59
+ """
60
+
61
+ _HOMEPAGE = "https://irvlab.cs.umn.edu/resources/suim-dataset"
62
+
63
+ _LICENSE = "mit"
64
+
65
+
66
+ class ExDark(datasets.GeneratorBasedBuilder):
67
+ """Semantic Segmentation of Underwater IMagery (SUIM) dataset"""
68
+
69
+ VERSION = datasets.Version("1.0.0")
70
+
71
+ BUILDER_CONFIGS = [
72
+ datasets.BuilderConfig(
73
+ name="suim",
74
+ version=VERSION,
75
+ description="Semantic Segmentation of Underwater IMagery (SUIM) dataset",
76
+ ),
77
+ ]
78
+
79
+ DEFAULT_CONFIG_NAME = "suim"
80
+
81
+ def _info(self):
82
+ return datasets.DatasetInfo(
83
+ description=_DESCRIPTION,
84
+ features=datasets.Features(
85
+ {
86
+ "img": datasets.Image(),
87
+ "mask": datasets.Image(),
88
+ }
89
+ ),
90
+ homepage=_HOMEPAGE,
91
+ license=_LICENSE,
92
+ citation=_CITATION,
93
+ )
94
+
95
+ def _split_generators(self, dl_manager):
96
+ data_dir = dl_manager.download_and_extract("SUIM.zip")
97
+ train_dir = os.path.join(data_dir, "SUIM", "train_val")
98
+ test_dir = os.path.join(data_dir, "SUIM", "TEST")
99
+
100
+ return [
101
+ datasets.SplitGenerator(
102
+ name=datasets.Split.TRAIN,
103
+ gen_kwargs={
104
+ "data_dir": train_dir,
105
+ "split": "train",
106
+ },
107
+ ),
108
+ datasets.SplitGenerator(
109
+ name=datasets.Split.TEST,
110
+ gen_kwargs={
111
+ "data_dir": test_dir,
112
+ "split": "test",
113
+ },
114
+ ),
115
+ ]
116
+
117
+ def _generate_examples(self, data_dir, split):
118
+ img_dir = os.path.join(data_dir, "images")
119
+ masks_dir = os.path.join(data_dir, "masks")
120
+ img_files = os.listdir(img_dir)
121
+
122
+ for idx, img_file in enumerate(img_files):
123
+ img_path = os.path.join(img_dir, img_file)
124
+ mask_path = os.path.join(
125
+ masks_dir,
126
+ img_file.replace(".jpg", ".bmp"),
127
+ )
128
+ record = {
129
+ "img": img_path,
130
+ "mask": mask_path,
131
+ }
132
+ yield idx, record