Dataset Viewer
Auto-converted to Parquet Duplicate
video_id
stringclasses
15 values
frame_index
int32
0
1.11k
image
imagewidth (px)
640
640
mask
imagewidth (px)
640
640
flow_rate_class
int8
1
10
official_split
stringclasses
9 values
FLIR0659
0
8
train
FLIR0659
1
8
train
FLIR0659
2
8
train
FLIR0659
3
8
train
FLIR0659
4
8
test
FLIR0659
5
8
train
FLIR0659
6
8
train
FLIR0659
7
8
validation
FLIR0659
8
8
train
FLIR0659
9
8
train
FLIR0659
10
8
train
FLIR0659
11
8
train
FLIR0659
12
8
train
FLIR0659
13
8
train
FLIR0659
14
8
train
FLIR0659
15
8
train
FLIR0659
16
8
train
FLIR0659
17
8
train
FLIR0659
18
8
train
FLIR0659
19
8
train
FLIR0659
20
8
train
FLIR0659
21
8
train
FLIR0659
22
8
train
FLIR0659
23
8
train
FLIR0659
24
8
train
FLIR0659
25
8
train
FLIR0659
26
8
validation
FLIR0659
27
8
train
FLIR0659
28
8
test
FLIR0659
29
8
validation
FLIR0659
30
8
test
FLIR0659
31
8
train
FLIR0659
32
8
test
FLIR0659
33
8
train
FLIR0659
34
8
train
FLIR0659
35
8
train
FLIR0659
36
8
train
FLIR0659
37
8
train
FLIR0659
38
8
train
FLIR0659
39
8
train
FLIR0659
40
8
train
FLIR0659
41
8
train
FLIR0659
42
8
train
FLIR0659
43
8
train
FLIR0659
44
8
train
FLIR0659
45
8
train
FLIR0659
46
8
train
FLIR0659
47
8
train
FLIR0659
48
8
train
FLIR0659
49
8
test
FLIR0659
50
8
validation
FLIR0659
51
8
train
FLIR0659
52
8
test
FLIR0659
53
8
train
FLIR0659
54
8
train
FLIR0659
55
8
train
FLIR0659
56
8
train
FLIR0659
57
8
train
FLIR0659
58
8
train
FLIR0659
59
8
train
FLIR0659
60
8
train
FLIR0659
61
8
train
FLIR0659
62
8
train
FLIR0659
63
8
train
FLIR0659
64
8
validation
FLIR0659
65
8
train
FLIR0659
66
8
train
FLIR0659
67
8
train
FLIR0659
68
8
train
FLIR0659
69
8
train
FLIR0659
70
8
train
FLIR0659
71
8
train
FLIR0659
72
8
train
FLIR0659
73
8
validation
FLIR0659
74
8
test
FLIR0659
75
8
train
FLIR0659
76
8
train
FLIR0659
77
8
test
FLIR0659
78
8
test
FLIR0659
79
8
train
FLIR0659
80
8
test
FLIR0659
81
8
train
FLIR0659
82
8
train
FLIR0659
83
8
train
FLIR0659
84
8
train
FLIR0659
85
8
train
FLIR0659
86
8
train
FLIR0659
87
8
train
FLIR0659
88
8
train
FLIR0659
89
8
validation
FLIR0659
90
8
train
FLIR0659
91
8
test
FLIR0659
92
8
train
FLIR0659
93
8
train
FLIR0659
94
8
train
FLIR0659
95
8
train
FLIR0659
96
8
train
FLIR0659
97
8
train
FLIR0659
98
8
train
FLIR0659
99
8
train
End of preview. Expand in Data Studio

Gasformer Dataset: MR and CR

Paper: arXiv | CVF | IEEE Xplore | Code: GitHub | Project: AI for Greener Livestock

Overview

This dataset has two subsets: MR (Controlled Methane Release) and CR (Dairy Cow Rumen Gas). It holds optical gas imaging frames with pixel masks of methane plumes. MR is a controlled methane release over an ice-block background at flow rates of 10 to 100 SCCM. CR is gas from dairy cow rumen samples.

Dataset Videos Labelled frames Resolution Camera Classes
MR, Controlled Methane Release 15 9,009 640x480 FLIR GF77 background + 10 flow-rate classes
CR, Dairy Cow Rumen Gas 1 340 640x480 FLIR GF77 background + gas

Data

Each parquet file is one video. Each row is one labelled frame, ordered by frame index.

Column Content
video_id video name, for example FLIR0659
frame_index frame number in the video, from 0
image original PNG, 3 channels
mask original PNG, 1 channel
flow_rate_class MR only. The gas class of the frame
official_split train, validation or test, or two joined by + (for example train+test)

Mask values: 0 is background. In MR, 1 to 10 is the flow-rate class (k x 10 SCCM), and mask > 0 gives the binary gas mask. In CR, 1 is gas.

Usage

from datasets import load_dataset

mr = load_dataset("toqi/Gasformer", "MR", split="all")
cr = load_dataset("toqi/Gasformer", "CR", split="all")
image, mask = mr[0]["image"], mr[0]["mask"]   # PIL images

python unpack_to_png.py <repo_dir> <out_dir> writes PNG folders and train.txt, val.txt, test.txt lists.

Splits

official_split.csv lists the train, validation or test split of every frame, as used in the Gasformer paper. We recommend splitting by video, because neighbouring frames are nearly identical.

License

License: CC BY-NC 4.0. Commercial use is not permitted.

Citation

If you use this dataset, you must cite the paper below.

@inproceedings{sarker2024gasformer,
  author    = {Sarker, Toqi Tahamid and Embaby, Mohamed G. and Ahmed, Khaled R. and AbuGhazaleh, Amer},
  title     = {Gasformer: A Transformer-based Architecture for Segmenting Methane Emissions from Livestock in Optical Gas Imaging},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
  month     = {June},
  year      = {2024},
  pages     = {5489--5497},
  doi       = {10.1109/CVPRW63382.2024.00558}
}

Acknowledgments

This work is supported by the project "Artificial Intelligence for Greener Livestock: Educational and Research", award no. 2022-70001-37404, from the U.S. Department of Agriculture's National Institute of Food and Agriculture.

Downloads last month
-

Models trained or fine-tuned on toqi/Gasformer

Paper for toqi/Gasformer