47 GB
7,107 files
Updated 20 days ago
Name
Size
001_CH4_PlumeComplex-485
002_CH4_PlumeComplex-486
003_CH4_PlumeComplex-487
004_CH4_PlumeComplex-496
006_CH4_PlumeComplex-491
008_CH4_PlumeComplex-495
009_CH4_PlumeComplex-742
010_CH4_PlumeComplex-497
012_CH4_PlumeComplex-500
013_CH4_PlumeComplex-502
014_CH4_PlumeComplex-498
015_CH4_PlumeComplex-743
016_CH4_PlumeComplex-744
017_CH4_PlumeComplex-503
018_CH4_PlumeComplex-507
019_CH4_PlumeComplex-508
020_CH4_PlumeComplex-506
021_CH4_PlumeComplex-505
022_CH4_PlumeComplex-510
024_CH4_PlumeComplex-514
025_CH4_PlumeComplex-512
027_CH4_PlumeComplex-517
028_CH4_PlumeComplex-516
029_CH4_PlumeComplex-519
030_CH4_PlumeComplex-520
031_CH4_PlumeComplex-522
032_CH4_PlumeComplex-523
035_CH4_PlumeComplex-530
037_CH4_PlumeComplex-527
038_CH4_PlumeComplex-524
039_CH4_PlumeComplex-525
040_CH4_PlumeComplex-533
041_CH4_PlumeComplex-534
042_CH4_PlumeComplex-535
043_CH4_PlumeComplex-536
044_CH4_PlumeComplex-537
045_CH4_PlumeComplex-538
046_CH4_PlumeComplex-539
047_CH4_PlumeComplex-540
050_CH4_PlumeComplex-526
051_CH4_PlumeComplex-915
052_CH4_PlumeComplex-916
053_CH4_PlumeComplex-917
055_CH4_PlumeComplex-919
056_CH4_PlumeComplex-920
059_CH4_PlumeComplex-924
060_CH4_PlumeComplex-925
061_CH4_PlumeComplex-923
064_CH4_PlumeComplex-928
068_CH4_PlumeComplex-544
069_CH4_PlumeComplex-545
070_CH4_PlumeComplex-740
071_CH4_PlumeComplex-546
072_CH4_PlumeComplex-547
073_CH4_PlumeComplex-548
074_CH4_PlumeComplex-554
075_CH4_PlumeComplex-555
076_CH4_PlumeComplex-556
086_CH4_PlumeComplex-550
087_CH4_PlumeComplex-551
091_CH4_PlumeComplex-566
095_CH4_PlumeComplex-572
096_CH4_PlumeComplex-574
098_CH4_PlumeComplex-576
099_CH4_PlumeComplex-578
100_CH4_PlumeComplex-579
102_CH4_PlumeComplex-582
103_CH4_PlumeComplex-583
104_CH4_PlumeComplex-584
105_CH4_PlumeComplex-586
106_CH4_PlumeComplex-587
107_CH4_PlumeComplex-590
108_CH4_PlumeComplex-591
110_CH4_PlumeComplex-593
111_CH4_PlumeComplex-594
116_CH4_PlumeComplex-739
118_CH4_PlumeComplex-601
120_CH4_PlumeComplex-603
121_CH4_PlumeComplex-604
122_CH4_PlumeComplex-599
123_CH4_PlumeComplex-605
125_CH4_PlumeComplex-610
126_CH4_PlumeComplex-611
127_CH4_PlumeComplex-614
130_CH4_PlumeComplex-617
131_CH4_PlumeComplex-618
132_CH4_PlumeComplex-621
133_CH4_PlumeComplex-622
134_CH4_PlumeComplex-623
137_CH4_PlumeComplex-624
140_CH4_PlumeComplex-627
141_CH4_PlumeComplex-628
142_CH4_PlumeComplex-629
143_CH4_PlumeComplex-630
145_CH4_PlumeComplex-635
146_CH4_PlumeComplex-640
148_CH4_PlumeComplex-636
149_CH4_PlumeComplex-637
151_CH4_PlumeComplex-643
.gitattributes257 kB
xet
README.md

OxHyperRealCH4 dataset

Author: Vít Růžička

One of the OxHyper datasets, for more details please check my main web at: https://previtus.github.io/

Fast data preview in: https://huggingface.co/datasets/previtus/OxHyperRealCH4/blob/main/dataset_exploration.ipynb

The "OxHyperRealCH4" dataset is a new mono-temporal hyperspectral dataset of real methane leak events. In total we have 279 large tiles in the training dataset, 91 in the validation dataset and 98 in the test dataset. These have reliable ground truth labels, which we calculated from the original labels provided on the EMIT data portal. We note that this dataset is smaller than the other ones and training a machine learning model from scratch could lead to overfitting and poor generalisation, however it can also be used for model fine-tuning and to test generalisation ability of models trained on other datasets. The total size of this data is 47 GB. It contains 86 spectral bands made of the RGB bands and ranges between 1573-1699nm and 2004-2478nm.

More info will be added upon publication, for now please refer to https://previtus.github.io/

More information about the NASA's sensor EMIT: https://earth.jpl.nasa.gov/emit/

File structure

Each folder represent a large tile (of 512x512px), in this case, every folder also contains a real methane leak event captured by the EMIT sensor. Rows in the attached csv's split these into train/val/test subsets.

Each folder contains these files:

  • B - all 86 bands in a georeferenced file (note: you'll also need the B.hdr file to load it, can be also inspected for example in QGIS)
  • B_EMIT_641nm.tif, B_EMIT_551nm.tif, B_EMIT_462nm.tif - The RGB bands, these were extracted from the larger file
  • label_v3_binary.tif - binary label (pixel values 0 or 1 for "no plume" or "plume" class)
  • B_magic30_tile.tif - computed matched filter product (using the iterative variant called mag1c)
  • info.json - saved information about the used source data (for example the link to the used EMIT_L2B_CH4PLM sample)
  • Note: for each georeferenced tif we also provide the .hdr file which is needed for loading and has details about the bands

Citation

Will be added.

Total size
47 GB
Files
7,107
Last updated
Oct 20
Pre-warmed CDN
US EU US EU

Contributors