---
license: mit
dataset_info:
  config_name: suim
  features:
  - name: img
    dtype: image
  - name: mask
    dtype: image
  splits:
  - name: train
    num_bytes: 511917
    num_examples: 1525
  - name: test
    num_bytes: 35774
    num_examples: 110
  download_size: 183261195
  dataset_size: 547691
task_categories:
- image-segmentation
---

The SUIM dataset is a dataset for semantic segmentation of underwater imagery.

The dataset consists of 1525 annotated images for training/validation and
110 samples for testing.

| Object category                  | Symbol | Grayscale value |
|----------------------------------|--------|-----------------|
| Background (waterbody)           | BW     | 0 (black)       |
| Human divers                     | HD     | 1 (blue)        |
| Aquatic plants and sea-grass     | PF     | 2 (green)       |
| Wrecks and ruins                 | WR     | 3 (sky)         |
| Robots (AUVs/ROVs/instruments)   | RO     | 4 (red)         |
| Reefs and invertebrates          | RI     | 5 (pink)        |
| Fish and vertebrates             | FV     | 6 (yellow)      |
| Sea-floor and rocks              | SR     | 7 (white)       |


For more information about the original SUIM dataset,
please visit the official dataset page:

https://irvlab.cs.umn.edu/resources/suim-dataset

Please refer to the original dataset source for any additional details,
citations, or specific usage guidelines provided by the dataset creators.