CQIS (Coffee Quality Inspection System)
Coffee Quality Inspection System
Company
UIN Raden Intan Lampung
Year
2026
Category
Mobile Computer Vision & Expert System
Role
Software & AI Engineer
Overview
The coffee industry in Indonesia requires accurate and efficient quality standardization to compete in the global market. Currently, the physical inspection of coffee beans is still done manually, which is prone to subjectivity bias and eye fatigue of the inspectors.
As the lead researcher and developer, my role was to design and build an end-to-end artificial intelligence system architecture. The resulting solution is CQIS (Coffee Quality Inspection System), a smart mobile application based on Computer Vision capable of detecting and classifying 20 types of physical defects in coffee beans according to the Indonesian National Standard (SNI 01-2907-2008). This project was developed independently by collaborating various technology stacks, integrating expertise in mobile software engineering, digital image processing, to server infrastructure management.
CQIS is equipped with real-time camera scanning features using the YOLOv11 Deep Learning model, dynamic size category mapping through a polygon ratio heuristic approach, an integrated dataset management system (Data Flywheel), and a web admin panel for inspection data monitoring. The main contribution of this system lies in its ability to convert descriptive SNI criteria into geometric mathematical parameters, enabling the AI machine to provide constant assessments, invariant to camera distance, and highly accurate.
The development of this application applies modern software engineering practices, including the separation of Frontend (Flutter) and Backend API (Python) architecture, the use of Supabase for Cloud Database with SQLite fallback support for offline mode, and deployment on a Linux VPS. Ultimately, CQIS is not just a prototype, but a ready-to-use digital product that lays the foundation for quality standardization digitalization, drives supply chain efficiency, and strengthens the objectivity of Indonesian coffee in the eyes of the world.
Tech Stack
Frontend & Mobile: Flutter (Dart), React, TypeScript
Backend & AI: Python, FastAPI, Ultralytics YOLOv11, OpenCV, PyTorch, NumPy
Database: SQLite, Supabase
Objectives
- Translate 20 physical defect criteria of SNI 01-2907-2008 into Computer Vision mathematical logic (YOLOv11) through polygon ratio calculations that are invariant to camera shooting distance.
- Provide a coffee inspection tool that completely eliminates human subjectivity bias to produce consistent, objective, and nationally standardized grading.
- Create an intuitive mobile application interface with a one-tap camera scanning feature so it is very easy to operate by farmers and lay quality inspectors in the field.
- Implement a Data Flywheel system on the Web Admin to export dataset history automatically, ensuring the AI model can continue to be trained to become smarter.
- Design a robust hybrid system architecture between Python, Cloud, and local memory fallback (SQLite) to ensure application reliability in all situations.
Project Galery
A glimpse into the project