This project was actually developed regarding from these issues and its solution:
- Inconvenient & Inefficient Workflow: The interns must waiting for manual logbook queue at guard post.
- Security Risks: The risk of missing logbook or even damaged logbook.
By implementing Face Recognition System, ATLAS can eliminate those issues and digitizes the attendance process, data integrity through any local databases (I'm using SQLite in this case) and providing a seamless user experience.
To improve accuracy in crowded environments, ATLAS utilizes an oval-shaped ROI layout. This ensures the system focuses only on the primary user in front of the camera, preventing "false triggers" from background movement or other people passing by.
The system calculates the similarity between faces using Euclidean distance algorithms. To ensure high reliability, I implemented a strict Confidence Threshold logic:
| Threshold Range | Status | Action |
|---|---|---|
| < 0.50 | ✅ Match Found | Attendance recorded successfully. |
| 0.50 - 0.65 | System prompts user to try a different angle. | |
| > 0.65 | ❌ Unknown | Face isn't recognized in the database. |
Note: A lower threshold indicates a higher degree of similarity between the detected face and the database record.
- Real-time Face Detection: Agile and accurate recognition using OpenCV, Pickle and Face Recognition Libraries.
- Sleek UI: Clean and understandable UI by providing oval layout in the middle and feedback UI made with Tkinter.
- Downloadable Logs: Effortless exportable logs and automatically converted to Excel format.
- Local Databases Integration: Secure interns log & profiles.
- Secure Exportable Logs: Protected system with hardcoded secrets for downloading those logs.
| Component | Technology |
|---|---|
| Language | Python |
| AI/ML Libraries | OpenCV, Face Recognition |
| Database | SQLite |
| GUI | Tkinter |
| Others | Pickle, Pillow, Numpy |
To ensure the system runs efficiently as I expected, so, I've designed these workflow for this project.
Unfortunately, these Project Architecture are still written in local language (Bahasa Indonesia), not in English.
