Face Detection Web v1.2.0 - Full Training Pipeline & Feedback System

Release v1.2.0: Full Training Pipeline & Feedback System 🚀

This release marks a significant milestone for the project, transitioning it into a fully reproducible and production-ready open-source system. It introduces a comprehensive training pipeline, a new user feedback API, and major DevOps enhancements.

🌟 New Features

• User Feedback API (/api/feedback): Added a robust endpoint to collect user ratings and comments, securely integrated with the MySQL database.
• Optimal Confidence Threshold: Updated the default inference confidence threshold to 0.32 across the system, strictly based on empirical F1-Curve benchmarks to maximize the balance between Precision and Recall.

🧠 Research & Reproducibility

• Complete Training Pipeline: Introduced the training/ directory.
• Jupyter Notebook: Added train_yolov12-face_widerface.ipynb allowing anyone to retrain the YOLOv12 models from scratch.
• Comprehensive Benchmarks: Included detailed evaluation reports (Loss, mAP, PR curves, F1 curves, and Confusion Matrices) for all 4 model variants (Nano, Small, Medium, Large).

🛠️ DevOps & Engineering Standards

• Security (.env): Implemented environment variables structure (.env.example) to securely manage database credentials without hardcoding them.
• Automated Quality Checks (lint.yml): Added GitHub Actions workflow for continuous integration to enforce code quality using Flake8, Black, and isort.
• Docker Optimization (.dockerignore): Significantly reduced Docker build context and image size by ignoring unnecessary local artifacts.
• Code Normalization (.gitattributes): Standardized line endings (LF) across all operating systems to prevent Git conflicts.
• Linter Config (.flake8): Added explicit rules for Python code styling.

🤝 Community & Open Source

• Sponsorship (FUNDING.yml): Enabled GitHub Sponsors/BuyMeACoffee integration to support the ongoing development of the project.

What's Changed

• Update SECURITY.md by @RevDra in #15
• Update README.md by @RevDra in #16
• Update index.html by @RevDra in #17
• Integrate Flake8 for minor code error detection by @RevDra in #18
• Update face_detection_yolov12.py by @RevDra in #19
• Create codeql.yml by @RevDra in #21
• Update pyproject.toml by @RevDra in #22
• Update web_app.py by @RevDra in #25
• Add environment variables for database and secret key by @RevDra in #26
• Lower confidence threshold for face detection by @RevDra in #27

Full Changelog: v1.1.0...v1.2.0

Face Detection Web v1.1.0 - Security Hardening & Docker Support

🚀 v1.1.0 Release Notes

This major update brings production-grade stability to the project, featuring full Docker support for easy deployment and critical security patches.

✨ New Features

• 🐳 Docker Support: Now includes Dockerfile and docker-compose.yml. You can deploy the entire stack with a single command: docker-compose up.
• 🎨 UI Polish: Updated the web interface footer with a modern layout, including a version badge and credits.

🔒 Security Enhancements (Critical)

• Path Traversal Fix: Implemented strict model path validation using pathlib.resolve() and werkzeug.secure_filename.
• Information Exposure Fix: API now returns generic error messages to clients while logging detailed tracebacks server-side.
• Security Policy: Added SECURITY.md with private reporting guidelines.

🛠 Technical Improvements

• Refactor: Cleaned up project structure for better maintainability.
• Dependencies: Locked versions in requirements.txt for stability.
• CI/CD: Integrated GitHub Actions for automated CodeQL analysis and Docker builds

📦 Quick Start with Docker

git clone https://github.com/RevDra/human_face_detection.git
cd human_face_detection
docker-compose up --build

---

Full Changelog: https://github.com/RevDra/human_face_detection/compare/v1.0.0...v1.1.0

Face Detection Web v1.0.0 - Initial Release

First official release featuring YOLOv12 architecture integration.

Key Features:
🚀 Support for YOLOv12 Nano, Small, Medium, Large.
📹 Real-time Webcam detection optimized for CPU.
🐳 Docker & Hugging Face Spaces support ready.

Assets included:

• Source code.
• Pre-trained YOLOv12 models (download below).