Shrimp disease detection
Using photos to spot signs of a viral disease in farmed shrimp.
- Status
- Co-authored research
- Year
- 2023
- Role
- One of three co-authors
- Area
- Machine learning
In plain English
Viral infections can sweep through shrimp farms quickly, so early warning matters to farmers. This study asked whether image models can spot signs of White Spot Syndrome Virus in shrimp photos. It is a research comparison and is not meant for diagnosing real stock.
How it works
- Collect labeled photos of healthy shrimp and shrimp showing signs of disease.
- Expand the dataset with image augmentation (creating varied copies of each photo).
- Train a simple CNN (a basic image-recognition network) as a baseline.
- Train EfficientNet-B0 (a compact pretrained image model) on the same data.
- Compare the two approaches and write up the findings in a paper.
What I did
I was one of three co-authors on the study. The notebooks and the paper live in my repository.
What I learned
In our comparison, EfficientNet-B0 gave the highest accuracy. A small, careful comparison taught me more than chasing a single high score would have.