---
license: cc-by-4.0
task_categories:
- image-classification
language:
- en
pretty_name: 'JAMBO, A Multi-Annotator Image Dataset for Benthic Habitat Classification '
size_categories:
- 1K<n<10K
---

The JAMBO dataset contains 3290 underwater images of the seabed captured by an ROV in temperate waters in the Jammer Bay area off the North West coast of Jutland, Denmark.
All the images have been annotated by six annotators to contain one of three classes: sand, stone, or bad. The three classes are defined as follows:
* **Sand** habitats are characterized as primarily sand or muddy sand with less than 5% clay and less than 30% cover of stones/boulders, vegetation, and mussel bed.
* **Stone** reef habitats are characterized by having more than 30% seabed cover of stones or boulders.
* **Bad** is a class used to label images that cannot be confidently annotated as containing one of the aforementioned habitat types by the annotator due to poor image quality, turbidity, or similar.

Each of the six annotators have labelled all the images (that is, six individual annotations are provided for each image), which allows for analyzing how inter-annotator disagreement can affect the performance of machine learning models.
Cross-validation splits and date-based splits are provided in the [jambo_splits_public.csv](jambo_splits_public.csv) file. Check out the starter notebook [howto_jambo.ipynb](howto_jambo.ipynb) to get started.

For more information about the dataset and baseline models, please see the paper to be presented the ECCV 2024 Computer Vision for Ecology (CV4E) Workshop: [Underwater Uncertainty: A Multi-Annotator Image Dataset for Benthic Habitat
Classification](https://vbn.aau.dk/ws/portalfiles/portal/738329849/JAMBO_ECCV_vbn.pdf)