---
task_categories:
- image-segmentation
- mask-generation
language:
- en
license: cc-by-4.0
dataset_info:
features:
- name: image
dtype: image
configs:
- config_name: example_images
data_files:
- split: group_1
path:
- metadata.csv
- "group_1/*.png"
- split: group_2
path:
- metadata.csv
- "group_2/*.png"
- split: group_3
path:
- metadata.csv
- "group_3/*.png"
- split: group_4
path:
- metadata.csv
- "group_4/*.png"
- split: group_5
path:
- metadata.csv
- "group_5/*.png"
- split: group_6
path:
- metadata.csv
- "group_6/*.png"
- split: group_7
path:
- metadata.csv
- "group_7/*.png"
- split: group_8
path:
- metadata.csv
- "group_8/*.png"
- split: group_9
path:
- metadata.csv
- "group_9/*.png"
- split: group_10
path:
- metadata.csv
- "group_10/*.png"
- split: group_11
path:
- metadata.csv
- "group_11/*.png"
- split: group_12
path:
- metadata.csv
- "group_12/*.png"
- split: group_13
path:
- metadata.csv
- "group_13/*.png"
- split: group_14
path:
- metadata.csv
- "group_14/*.png"
- split: group_15
path:
- metadata.csv
- "group_15/*.png"
- split: group_16
path:
- metadata.csv
- "group_16/*.png"
- split: group_17
path:
- metadata.csv
- "group_17/*.png"
- split: group_18
path:
- metadata.csv
- "group_18/*.png"
- split: group_19
path:
- metadata.csv
- "group_19/*.png"
- split: group_20
path:
- metadata.csv
- "group_20/*.png"
- split: group_21
path:
- metadata.csv
- "group_21/*.png"
- split: group_22
path:
- metadata.csv
- "group_22/*.png"
- split: group_23
path:
- metadata.csv
- "group_23/*.png"
- split: group_24
path:
- metadata.csv
- "group_24/*.png"
- split: group_25
path:
- metadata.csv
- "group_25/*.png"
#
---
The **AUTOFISH** dataset comprises 1500 high-quality images of fish on a conveyor belt. It features 454 unique fish with class labels, IDs, manual length measurements,
and a total of 18,160 instance segmentation masks.
The fish are partitioned into 25 groups, with 14 to 24 fish in each group. Each fish only appears in one group, making it easy to create training splits. The
number of fish and distribution of species in each group were pseudo-randomly selected to mimic real-world scenarios.
Every group is partitioned into three subsets: *Set1*, *Set2*, and *All*. *Set1* and *Set2* contain half of the fish each, and none of the
fish overlap or touch each other. *All* contains all the fish from the group, purposely placed in positions with high overlap. Every group directory contains 20 images for
each set, where variation is introduced by changing the position and orientation of the fish. Exactly half of every set is with the fish on their one side, while the other
half has the fish flipped. The following figures display some examples with overlaid annotations:
| | |
|----------|----------|
| | |
| | |
The available classes are:
- Cod
- Haddock
- Whiting
- Hake
- Horse mackerel
- Other
Other information contained in the annotations:
- Segmentation masks
- Bounding boxes
- Lengths
- Unique fish IDs
- 'Side up' referring to the side of the fish that is visible
In addition to all the labeled data, two high-overlap
unlabeled groups, as well as camera calibration images are included.
You can load this dataset with a default split configuration using the datasets library
```python
dataset = datasets.load_dataset('vapaau/autofish', revision='script', trust_remote_code=True)
```
If you use this dataset for your work, please cite:
```yaml
Citation coming soon
```