ShrimpAI
Automated Disease Detection in Pacific White Shrimp
Company
UIN Raden Intan Lampung
Year
2026
Category
Artificial Intelligence, Computer Vision & Progressive Web Application (PWA)
Role
AI & Fullstack Web Developer
Overview
To support aquaculture farmers and technicians in preventing catastrophic harvest failures caused by shrimp pathogens, the ShrimpAI platform was developed as an AI-powered Progressive Web Application (PWA) for automated disease detection in Pacific White Shrimp (Litopenaeus vannamei). As the AI & Web Developer, I was responsible for end-to-end design and implementation, from training the deep learning computer vision model to engineering the fullstack web architecture and responsive mobile-first interface.
The platform enables farmers to capture or upload shrimp photos on-field and receive real-time diagnostic bounding boxes, health status classifications (Healthy, Black Gill / BG, White Spot Syndrome Virus / WSSV, or Co-infections), confidence scores, and tailored clinical treatment recommendations. To support comprehensive farm management, the system features historical diagnosis tracking, metric visualizations, and automated single-item and recap PDF report generation for pond monitoring records.
My key contributions centered on fine-tuning a custom YOLOv11 model (trained on 3,301 images across 200 epochs on a Tesla T4 GPU, achieving 99.2% mAP), building a high-performance Flask backend with role-based access control (RBAC), implementing PWA capabilities (Service Worker caching and direct on-device camera integration), and crafting an intuitive, responsive user interface styled with Tailwind CSS.
Tech Stack
- AI & Computer Vision: Python, Ultralytics YOLOv11, PyTorch, OpenCV, Pillow
- Backend & Database: Flask, Flask-SQLAlchemy, Flask-Login, Flask-Migrate, MySQL
- Frontend & PWA: Tailwind CSS, Vanilla JavaScript, HTML5, Service Worker, Web App Manifest, Font Awesome
- Reporting & Export: xhtml2pdf (
pisa)
Objectives
- Empower shrimp farmers with instant, automated on-field disease detection using deep learning computer vision to minimize harvest losses.
- Deliver an accessible, high-performance Progressive Web Application (PWA) optimized for mobile devices and field environments with native camera capture.
- Provide immediate, actionable bio-security recommendations and water treatment guidelines based on specific detected pathology.
- Facilitate continuous pond health auditing through digital inspection logs and downloadable PDF summary reports.
- Maintain a secure, scalable architecture with role-based access control (Petambak vs Administrator) and a dynamic disease knowledge base.
Project Galery
A glimpse into the project