Related Experiment Video
Updated: Jun 19, 2026
10:56
Long-term Behavioral Tracking of Freely Swimming Weakly Electric Fish
Published on: March 6, 2014
A Freshwater Fish Dataset for Visual Recognition with Manually Localized ROIs and SAM-Derived Instance Masks
Shakib Absar1, Iftekhar Ahmed1, Md Istiaque Khalique1
1Interdisciplinary Computer Science (InteX) Research Lab, Sylhet, 3100, Bangladesh.
Scientific Data
|June 17, 2026
Summary
The SylFishBD dataset provides 9,075 real-world images of 9 fish species in Bangladesh, enabling automated identification and freshness assessment in markets. This benchmark supports computer vision for aquaculture and trade transparency.
Area of Science:
- Computer Vision
- Aquaculture Technology
- Data Science
Background:
- Automated fish species identification is crucial for Bangladesh's aquaculture and markets.
- Existing datasets lack real-world complexity, limiting model deployment.
Purpose of the Study:
- Introduce the SylFishBD dataset for automated fish identification in Bangladesh.
- Provide a benchmark dataset reflecting authentic market conditions.
Main Methods:
- Collected 9,075 high-resolution images of 9 freshwater fish species over 7 months.
- Captured images in uncontrolled market settings with diverse lighting and backgrounds.
- Annotated images with bounding boxes and segmentation masks using the Segment Anything Model (SAM).
Main Results:
- The SylFishBD dataset features 9,075 standardized 500x500px images.
- Images represent 9 prevalent freshwater species under realistic market conditions.
- Dataset includes precise annotations for computer vision tasks.
Conclusions:
- SylFishBD bridges the gap between controlled datasets and real-world applications.
- The dataset facilitates the development of deployable models for aquaculture, trade, and regulation.
- Enables advancements in automated species recognition and freshness assessment.
