Yolo Vision Hub

A comprehensive computer vision project featuring face recognition, object detection, and image processing algorithms. This project integrates multiple vision tasks into a unified Streamlit web application.

📋 Project Overview

This repository contains three main tasks, each demonstrating different computer vision techniques:

1. Face Recognition System (Folder 1/)
2. YOLO Object Detection (Folder 2/)
3. Image Processing Algorithms (Folder 3/)

🚀 Features

Task 1: Face Recognition System (1/)

A complete face recognition pipeline using OpenCV's YuNet face detector and SFace face recognizer, combined with an SVM classifier.

Components:

• Buoc1 (1/Buoc1/get_face.py): Face detection and extraction from video stream

  • Real-time face detection from IP camera
  • Automatic face alignment and cropping
  • Saves detected faces to dataset folder

• Buoc2 (1/Buoc2/Training.py): Model training

  • Extracts 128-dimensional face embeddings using SFace
  • Trains LinearSVC classifier on face embeddings
  • Saves trained model (svc.pkl) for recognition

• Buoc3 (1/Buoc3/predict.py): Real-time face recognition

  • Live face recognition from video stream
  • Displays recognized person names in real-time
  • Visualizes face landmarks and bounding boxes

Models Used:

• face_detection_yunet_2023mar.onnx: Face detection model
• face_recognition_sface_2021dec.onnx: Face recognition/embedding model
• svc.pkl: Trained SVM classifier

Dataset:

• Face images organized by person in 1/image/ directory
• Currently includes: DaiLong, MinhHieu, MinhHoang, VanThang, caothang

Usage:

# Step 1: Collect face images
cd 1/Buoc1
python get_face.py --video http://192.168.0.100:8080/video

# Step 2: Train the classifier
cd ../Buoc2
python Training.py

# Step 3: Run face recognition
cd ../Buoc3
python predict.py --video http://192.168.0.101:8080/video

---

Task 2: YOLO Object Detection (2/)

YOLO v8-based object detection system for fruit recognition.

Features:

• Training Script (2/train_yolo.py): Custom YOLO model training
• GUI Application (2/yolo_detector_gui.py): Tkinter-based desktop application
  • Image upload and prediction
  • Real-time object detection with bounding boxes
  • Confidence score display

Classes Detected:

1. SauRieng (Durian)
2. Tao (Apple)
3. ThanhLong (Dragon Fruit)
4. Chuoi (Banana)
5. Kiwi

Dataset:

• Training data: 2/data/TraiCayx5-640x640_OK/
  • Train set: 288 images
  • Validation set: 72 images
• Raw images: 2/data/TraiCayScratch/
• Model: 2/model/best.onnx

Usage:

# Train YOLO model
cd 2
python train_yolo.py

# Run GUI application
python yolo_detector_gui.py

---

Task 3: Image Processing Algorithms (3/)

Comprehensive collection of image processing algorithms organized into three chapters.

Chapter 3: Basic Image Enhancement (3/chapter3.py)

• Point Operations:

  • Negative transformation
  • Logarithmic transformation
  • Power-law (gamma) transformation
  • Piecewise linear transformation

• Histogram Processing:

  • Histogram visualization
  • Histogram equalization (grayscale & color)
  • Local histogram equalization
  • Histogram statistics-based enhancement

• Spatial Filtering:

  • Smoothing filters (Box, Gaussian)
  • Median filter
  • Sharpening filters (Laplacian, Unsharp masking)
  • Gradient computation

Chapter 4: Frequency Domain Filtering (3/chapter4.py)

• FFT spectrum visualization
• Frequency domain filtering
• Moire pattern removal
• Motion blur simulation
• Interference removal

Chapter 9: Morphological Operations (3/chapter9.py)

• Erosion
• Dilation
• Boundary extraction
• Contour detection and visualization

Usage:

cd 3
python test_c3_4_9.py

---

🌐 Streamlit Web Application

A unified web interface (streamlit/app.py) that integrates all three tasks:

Features:

1. Face Recognition (Nhận diện khuôn mặt)

   • Image upload
   • Video upload
   • Real-time webcam streaming (WebRTC)

2. Object Detection (Nhận diện đối tượng)

   • Fruit detection using YOLO model
   • Real-time prediction with confidence scores

3. Image Processing (Xử lý ảnh số)

   • Interactive UI for all Chapter 3, 4, and 9 operations
   • Before/after image comparison
   • Export processed images

4. Handwriting OCR (Đọc chữ viết tay)

   • Handwritten text recognition using LSTM-based model

Running the Streamlit App:

cd streamlit
pip install -r requirements.txt
streamlit run app.py

The app will be available at http://localhost:8501

---

📁 Project Structure

Yolo-Vision-Hub/
├── 1/                          # Face Recognition Task
│   ├── Buoc1/                  # Face collection
│   │   ├── get_face.py
│   │   └── get_face - Copy.py
│   ├── Buoc2/                  # Training
│   │   └── Training.py
│   ├── Buoc3/                  # Prediction
│   │   └── predict.py
│   ├── image/                  # Face dataset
│   │   ├── caothang/
│   │   ├── DaiLong/
│   │   ├── MinhHieu/
│   │   ├── MinhHoang/
│   │   └── VanThang/
│   └── model/                  # Pre-trained models
│       ├── face_detection_yunet_2023mar.onnx
│       ├── face_recognition_sface_2021dec.onnx
│       └── svc.pkl
│
├── 2/                          # YOLO Object Detection
│   ├── data/                   # Training dataset
│   │   ├── TraiCayScratch/     # Raw images
│   │   └── TraiCayx5-640x640_OK/  # Processed dataset
│   ├── model/                  # YOLO model
│   │   └── best.onnx
│   ├── train_yolo.py          # Training script
│   ├── train_yolo_v8n(2).ipynb
│   ├── yolo_detector_gui.py   # GUI application
│   └── yolo2.py
│
├── 3/                          # Image Processing
│   ├── chapter3.py            # Basic enhancement
│   ├── chapter4.py            # Frequency domain
│   ├── chapter9.py            # Morphology
│   ├── test_c3_4_9.py        # Test script
│   └── image_test/            # Test images
│
├── 4/                          # OCR Model Training
│   ├── data_test/
│   ├── model/
│   │   └── best_model.h5
│   └── train.ipynb
│
└── streamlit/                  # Web application
    ├── app.py                 # Main Streamlit app
    ├── chapter3.py, chapter4.py, chapter9.py
    ├── face_recognition_utils.py
    ├── video_processor.py
    ├── stream_processor.py
    ├── ocr_utils.py
    ├── requirements.txt
    └── model/                 # All models
        ├── best_model.h5      # OCR model
        ├── best.onnx          # YOLO model
        ├── face_detection_yunet_2023mar.onnx
        ├── face_recognition_sface_2021dec.onnx
        └── svc.pkl            # Face recognition classifier

---

🛠️ Dependencies

Core Libraries:

• opencv-python / opencv-contrib-python - Computer vision operations
• numpy - Numerical operations
• scikit-learn - Machine learning (SVM classifier)
• ultralytics - YOLO v8 implementation
• joblib - Model serialization

For Streamlit App:

• streamlit - Web application framework
• streamlit-webrtc - Real-time video streaming
• keras / tensorflow - Deep learning (OCR model)

Installation:

# Install core dependencies
pip install opencv-python opencv-contrib-python numpy scikit-learn ultralytics joblib

# Install Streamlit dependencies
cd streamlit
pip install -r requirements.txt

---

📝 Usage Examples

Face Recognition Pipeline:

1. Collect face data: Run 1/Buoc1/get_face.py with your IP camera
2. Train model: Execute 1/Buoc2/Training.py to train the SVM classifier
3. Recognize faces: Use 1/Buoc3/predict.py for real-time recognition

YOLO Fruit Detection:

1. Train model: Use 2/train_yolo.py or the Jupyter notebook
2. Run GUI: Launch 2/yolo_detector_gui.py for desktop application
3. Use in Streamlit: Access through the web interface

Image Processing:

1. Import functions from chapter3.py, chapter4.py, or chapter9.py
2. Use the Streamlit web interface for interactive processing
3. Or run test_c3_4_9.py for batch processing

---

🎯 Key Technologies

• Face Recognition: OpenCV DNN (YuNet + SFace) + SVM
• Object Detection: YOLO v8 (Ultralytics)
• Image Processing: OpenCV + NumPy
• Web Interface: Streamlit + WebRTC
• OCR: LSTM-based neural network (Keras/TensorFlow)

---

📄 License

This project is for educational and research purposes.

---

👤 Author

Demo by HieuDuong

---

🔗 Model Sources

• YuNet Face Detector: OpenCV Zoo
• SFace Face Recognizer: OpenCV Zoo
• YOLO v8: Ultralytics

---

📌 Notes

• IP camera URLs need to be configured in the scripts
• Model paths may need adjustment based on your system
• Ensure all model files are present in the respective model/ directories
• For best results, use high-quality images and proper lighting conditions