fish-gang/deepscan-dataset

Tropical Reef Fish Dataset

A dataset of tropical reef fish images for species-level image classification, collected from iNaturalist.

Dataset Description

Images were scraped from iNaturalist using research-grade observations with CC0 or CC-BY licenses. The dataset covers 12 species across 9 families, targeting fish commonly encountered by snorkelers at depths of 0-10 meters in tropical reef environments.

This dataset was created as part of a bachelor thesis at ZHAW School of Engineering (Institute of Embedded Systems, InES) with the goal of training a lightweight image classifier for fully offline deployment on mobile devices.

Species

Scientific Name	Common Name	Family	Images
Acanthurus coeruleus	Atlantic Blue Tang	Acanthuridae	500
Naso unicornis	Bluespine Unicornfish	Acanthuridae	500
Scarus ghobban	Blue-barred Parrotfish	Labridae	500
Thalassoma lunare	Moon Wrasse	Labridae	500
Amphiprion ocellaris	Ocellaris Anemonefish	Pomacentridae	356
Chromis viridis	Blue-green Chromis	Pomacentridae	413
Chaetodon lunula	Raccoon Butterflyfish	Chaetodontidae	500
Pomacanthus imperator	Emperor Angelfish	Pomacanthidae	291
Rhinecanthus aculeatus	Lagoon Triggerfish	Balistidae	500
Pterois volitans	Red Lionfish	Scorpaenidae	500
Arothron meleagris	Guineafowl Puffer	Tetraodontidae	499
Carcharhinus melanopterus	Blacktip Reef Shark	Carcharhinidae	468

Total: 5,527 images across 12 species

Data Collection

• Source: iNaturalist API (research-grade observations only)
• Licenses: CC0 and CC-BY only, validated at the individual photo level
• Image resolution: Original resolution as provided by iNaturalist
• Target depth range: 0–10m (snorkeling depth)
• Contributors: Leisure snorkelers and underwater photographers worldwide

Structure

data/
  <family_scientific_name>/
    <species_scientific_name>/
      <species_scientific_name>_0000.jpg
      ...
dataset.json

Metadata

Each image entry in the data array of dataset.json contains:

• species_taxon_id — iNaturalist taxon ID for the species (links to labels array)
• file — relative path to the image file
• observation_id / observation_uri — link back to the original iNaturalist observation
• photo_id / photo_url_original — original photo reference on iNaturalist
• photo_license — per-photo license (cc0 or cc-by)
• photo_attribution — attribution string as required by the license

Species and family metadata is stored separately in the labels array, accessed via species_taxon_id keys:

• species_taxon_id / family_taxon_id — iNaturalist taxon IDs
• family_common_name / family_scientific_name — family-level classification
• species_common_name / species_scientific_name — species-level classification
• image_count — number of images for this species in the dataset

Intended Use

This dataset is intended for training lightweight convolutional neural networks for species-level fish classification on mobile devices. It is not exhaustive of all tropical reef fish species and is not intended for scientific surveys or biodiversity assessments.

Limitations

• Moderate class imbalance (291–500 images per species)
• No bounding box annotations — classification labels only
• Image quality varies (motion blur, partial occlusion, waterproof housing distortion)

Attribution

This dataset contains images contributed by iNaturalist observers under CC0 and CC-BY licenses. Full per-image attribution is available in dataset.json under the photo_attribution field. Please credit original observers when redistributing or publishing work derived from this dataset.

License

Dataset metadata and structure: CC-BY 4.0

Individual images: see photo_license field in dataset.json (CC0 or CC-BY)