Fish freshness with computer vision (dissertation)
My final-year dissertation: from a photo, it decides whether a fish is fresh or not.
Facts
- 98.19%accuracy on the test photos
- Python
- PyTorch
- torchvision
- scikit-learn
- Gradio
My final-year dissertation in Artificial Intelligence (De Montfort University). From a photo, the system says whether a fish is fresh or not.
I compared four pretrained neural networks (EfficientNetB0, EfficientNetB4, MobileNetV2 and ResNet50) fine-tuned for the problem, and a model of my own that combines three of them. The best one, EfficientNetB0, gets 98.19% of the test photos right. It can be used from a small web app where you upload a photo.
Built with Python, PyTorch and Gradio.
Contact
Send me an email.
BSc AI · First ClassDe Montfort University
Python · TypeScript · LangGraphFastAPI · Astro · Three.js
Vigo, SpainSpanish · Fluent English