Datasets:
metadata
license: cc-by-nc-4.0
task_categories:
- image-classification
paperswithcode_id: uvp6net
pretty_name: 'UVP6Net: plankton images captured with the UVP6'
dataset_info:
features:
- name: image
dtype: image
- name: label
dtype:
class_label:
names:
'0': Actinopterygii
'1': Amphipoda
'2': Annelida
'3': Appendicularia
'4': Aulacanthidae
'5': Aulatractus
'6': Aulosphaeridae
'7': Calanidae
'8': Calanoida
'9': Chaetognatha
'10': Coelodendridae
'11': Collodaria
'12': Copepoda
'13': Creseis acicula
'14': Ctenophora
'15': Echinodermata
'16': Eumalacostraca
'17': Foraminifera
'18': Hyperiidea
'19': Narcomedusae
'20': Ostracoda
'21': Phyllodocida
'22': Rhizaria
'23': Salpida
'24': Siphonophorae
'25': Swima
'26': Thecosomata
'27': Trachymedusae
'28': artefact
'29': chain<Salpida
'30': cloud
'31': copepoda_eggs
'32': crystal
'33': darksphere
'34': dead<house
'35': detritus
'36': fiber
'37': filament
'38': fuzzy
'39': like<Copepoda
'40': like<Rhizaria
'41': other<Cnidaria
'42': other<living
'43': puff
'44': reflection
'45': small<Cnidaria
'46': solitaryblack
'47': solitaryglobule
'48': spiky<Acantharia
'49': spiky<Coelodendridae
'50': star<Acantharia
'51': t004
'52': t011
'53': tuff
splits:
- name: train
num_bytes: 12409074296.238
num_examples: 634459
download_size: 607347956
dataset_size: 12409074296.238
description: >
Plankton was imaged with UVP6 in contrasted oceanic regions. The full images
were processed by the
UVP6 firmware and the regions of interest (ROIs) around each individual
object were recorded. A set
of associated features were measured on the objects (see Picheral et al.
2021, doi:10.1002/lom3.10475,
for more information). All objects were classified by a limited number of
operators into 110 different
classes using the web application EcoTaxa (http://ecotaxa.obs-vlfr.fr). The
following dataset corresponds
to the 634 459 objects that have an area superior to 73 pixels (equivalent
spherical diameter of 9.8 pixels,
corresponding to the default size limit of 620µm in the UVP6 configuration).
The different files provide
information about the features of the objects, their taxonomic
identification as well as the raw images.
For the purpose of training machine learning classifiers, the images in each
class were split into training,
validation, and test sets, with proportions 70%, 15% and 15%.
An additional folder is provided, which includes the subset of images used
to train the unique embedded
classification model of the UVP6 actually deployed on the NKE CTS5 floats
(10.5281/zenodo.10694203). These
images correspond to UVP6Net objects filtered to retain only those with a
size of 79 pixels to fit with the
645µm class from EcoPart, resulting in a total of 595,595 objects. The
taxonomic identification was also made
coarser (from 110 classes to 20) to ensure adequate performance of the
classification model on power-constrained
hardware. Images in this subset display objects as shades of grey/white on a
black background.
dataset_name: 'UVP6Net: plankton images captured with the UVP6'
citation: |-
@article{dataset:uvp6net,
title = {UVP6Net: plankton images captured with the UVP6},
author = {Picheral, Marc and Jalabert, Laetitia and Motreuil, Solène
and Courchet, Lucas and Carray-Counil, Louis and Ricour, Florian
and Panaiotis, Thelma and Petit, Flavien and Elineau, Amanda},
year = 2024,
journal = {SEANOE},
doi = {10.17882/101948},
affiliation = {Sorbonne Université, CNRS, Laboratoire d'Océanographie de Villefanche, LOV, 06230 Villefranche-sur-mer, France.
Sorbonne Université, Institut de la Mer de Villefranche, IMEV, 06230 Villefranche-sur-Mer, France.
Freshwater and OCeanic science Unit of reSearch (FOCUS), University of Liège, Liège, Belgium.}
}
homepage: https://www.seanoe.org/data/00908/101948/
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
dataset_description: >
Plankton was imaged with UVP6 in contrasted oceanic regions. The full images
were processed by the
UVP6 firmware and the regions of interest (ROIs) around each individual object
were recorded. A set
of associated features were measured on the objects (see Picheral et al. 2021,
doi:10.1002/lom3.10475,
for more information). All objects were classified by a limited number of
operators into 110 different
classes using the web application EcoTaxa (http://ecotaxa.obs-vlfr.fr). The
following dataset corresponds
to the 634 459 objects that have an area superior to 73 pixels (equivalent
spherical diameter of 9.8 pixels,
corresponding to the default size limit of 620µm in the UVP6 configuration).
The different files provide
information about the features of the objects, their taxonomic identification
as well as the raw images.
For the purpose of training machine learning classifiers, the images in each
class were split into training,
validation, and test sets, with proportions 70%, 15% and 15%.
An additional folder is provided, which includes the subset of images used to
train the unique embedded
classification model of the UVP6 actually deployed on the NKE CTS5 floats
(10.5281/zenodo.10694203). These
images correspond to UVP6Net objects filtered to retain only those with a size
of 79 pixels to fit with the
645µm class from EcoPart, resulting in a total of 595,595 objects. The
taxonomic identification was also made
coarser (from 110 classes to 20) to ensure adequate performance of the
classification model on power-constrained
hardware. Images in this subset display objects as shades of grey/white on a
black background.
source_url: https://www.seanoe.org/data/00908/101948/
citation_bibtex: |-
@article{dataset:uvp6net,
title = {UVP6Net: plankton images captured with the UVP6},
author = {Picheral, Marc and Jalabert, Laetitia and Motreuil, Solène
and Courchet, Lucas and Carray-Counil, Louis and Ricour, Florian
and Panaiotis, Thelma and Petit, Flavien and Elineau, Amanda},
year = 2024,
journal = {SEANOE},
doi = {10.17882/101948},
affiliation = {Sorbonne Université, CNRS, Laboratoire d'Océanographie de Villefanche, LOV, 06230 Villefranche-sur-mer, France.
Sorbonne Université, Institut de la Mer de Villefranche, IMEV, 06230 Villefranche-sur-Mer, France.
Freshwater and OCeanic science Unit of reSearch (FOCUS), University of Liège, Liège, Belgium.}
}
citation_apa: >
Picheral, M., Jalabert, L., Motreuil, S., Courchet, L., Carray-Counil, L.,
Ricour, F., Panaiotis, T., Petit, F., &
Elineau, A. (2024). UVP6Net: plankton images captured with the UVP6. SEANOE.
https://doi.org/10.17882/101948
hf_dataset_name: uvp6net
hf_org_name: project-oceania
report_markdown: >
**Samples per class for split `train`**
```──────────────────────── Label histogram for train split ─────────────────────────
0: Actinopterygii 110.00
1: Amphipoda 182.00
2: Annelida 247.00
3: Appendicularia 144.00
4: Aulacanthidae 295.00
5: Aulatractus 111.00
6: Aulosphaeridae 364.00
7: Calanidae 877.00
8: Calanoida ▇ 13506.00
9: Chaetognatha 572.00
10: Coelodendridae 211.00
11: Collodaria 658.00
12: Copepoda ▇ 6119.00
13: Creseis acicula 153.00
14: Ctenophora 204.00
15: Echinodermata 104.00
16: Eumalacostraca 1311.00
17: Foraminifera 188.00
18: Hyperiidea 90.00
19: Narcomedusae 113.00
20: Ostracoda 541.00
21: Phyllodocida 129.00
22: Rhizaria 2202.00
23: Salpida 207.00
24: Siphonophorae 122.00
25: Swima 236.00
26: Thecosomata 185.00
27: Trachymedusae 87.00
28: artefact ▇▇▇ 30139.00
29: chain<Salpida 153.00
30: cloud 242.00
31: copepoda_eggs 195.00
32: crystal 3770.00
33: darksphere 340.00
34: dead<house 909.00
35: detritus ▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇
508817.00
36: fiber ▇▇▇ 32592.00
37: filament 3306.00
38: fuzzy 184.00
39: like<Copepoda 2717.00
40: like<Rhizaria 545.00
41: other<Cnidaria 412.00
42: other<living 1583.00
43: puff ▇ 7427.00
44: reflection 4593.00
45: small<Cnidaria 927.00
46: solitaryblack 149.00
47: solitaryglobule 247.00
48: spiky<Acantharia 201.00
49: spiky<Coelodendridae 109.00
50: star<Acantharia 411.00
51: t004 249.00
52: t011 111.00
53: tuff 4863.00
```
dataset_means: '[0.9460734817311702]'
dataset_stds: '[0.1208790275683871]'
Dataset UVP6Net: plankton images captured with the UVP6
Plankton was imaged with UVP6 in contrasted oceanic regions. The full images were processed by the UVP6 firmware and the regions of interest (ROIs) around each individual object were recorded. A set of associated features were measured on the objects (see Picheral et al. 2021, doi:10.1002/lom3.10475, for more information). All objects were classified by a limited number of operators into 110 different classes using the web application EcoTaxa (http://ecotaxa.obs-vlfr.fr). The following dataset corresponds to the 634 459 objects that have an area superior to 73 pixels (equivalent spherical diameter of 9.8 pixels, corresponding to the default size limit of 620µm in the UVP6 configuration). The different files provide information about the features of the objects, their taxonomic identification as well as the raw images. For the purpose of training machine learning classifiers, the images in each class were split into training, validation, and test sets, with proportions 70%, 15% and 15%.
An additional folder is provided, which includes the subset of images used to train the unique embedded classification model of the UVP6 actually deployed on the NKE CTS5 floats (10.5281/zenodo.10694203). These images correspond to UVP6Net objects filtered to retain only those with a size of 79 pixels to fit with the 645µm class from EcoPart, resulting in a total of 595,595 objects. The taxonomic identification was also made coarser (from 110 classes to 20) to ensure adequate performance of the classification model on power-constrained hardware. Images in this subset display objects as shades of grey/white on a black background.
- Original dataset available online at: https://www.seanoe.org/data/00908/101948/.
- Original dataset license: <cc-by-nc-4.0>.
Details
- train split means (RGB): [0.9460734817311702]
- train split standard deviations (RGB): [0.1208790275683871]
Samples per class for split train
0: Actinopterygii 110.00
1: Amphipoda 182.00
2: Annelida 247.00
3: Appendicularia 144.00
4: Aulacanthidae 295.00
5: Aulatractus 111.00
6: Aulosphaeridae 364.00
7: Calanidae 877.00
8: Calanoida ▇ 13506.00
9: Chaetognatha 572.00
10: Coelodendridae 211.00
11: Collodaria 658.00
12: Copepoda ▇ 6119.00
13: Creseis acicula 153.00
14: Ctenophora 204.00
15: Echinodermata 104.00
16: Eumalacostraca 1311.00
17: Foraminifera 188.00
18: Hyperiidea 90.00
19: Narcomedusae 113.00
20: Ostracoda 541.00
21: Phyllodocida 129.00
22: Rhizaria 2202.00
23: Salpida 207.00
24: Siphonophorae 122.00
25: Swima 236.00
26: Thecosomata 185.00
27: Trachymedusae 87.00
28: artefact ▇▇▇ 30139.00
29: chain<Salpida 153.00
30: cloud 242.00
31: copepoda_eggs 195.00
32: crystal 3770.00
33: darksphere 340.00
34: dead<house 909.00
35: detritus ▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 508817.00
36: fiber ▇▇▇ 32592.00
37: filament 3306.00
38: fuzzy 184.00
39: like<Copepoda 2717.00
40: like<Rhizaria 545.00
41: other<Cnidaria 412.00
42: other<living 1583.00
43: puff ▇ 7427.00
44: reflection 4593.00
45: small<Cnidaria 927.00
46: solitaryblack 149.00
47: solitaryglobule 247.00
48: spiky<Acantharia 201.00
49: spiky<Coelodendridae 109.00
50: star<Acantharia 411.00
51: t004 249.00
52: t011 111.00
53: tuff 4863.00
Reference
Picheral, M., Jalabert, L., Motreuil, S., Courchet, L., Carray-Counil, L., Ricour, F., Panaiotis, T., Petit, F., & Elineau, A. (2024). UVP6Net: plankton images captured with the UVP6. SEANOE. https://doi.org/10.17882/101948
BibTEX
@article{dataset:uvp6net,
title = {UVP6Net: plankton images captured with the UVP6},
author = {Picheral, Marc and Jalabert, Laetitia and Motreuil, Solène
and Courchet, Lucas and Carray-Counil, Louis and Ricour, Florian
and Panaiotis, Thelma and Petit, Flavien and Elineau, Amanda},
year = 2024,
journal = {SEANOE},
doi = {10.17882/101948},
affiliation = {Sorbonne Université, CNRS, Laboratoire d'Océanographie de Villefanche, LOV, 06230 Villefranche-sur-mer, France.
Sorbonne Université, Institut de la Mer de Villefranche, IMEV, 06230 Villefranche-sur-Mer, France.
Freshwater and OCeanic science Unit of reSearch (FOCUS), University of Liège, Liège, Belgium.}
}
Usage
from datasets import load_dataset
dataset = load_dataset("project-oceania/uvp6net")