ENZYME

The Open-Source Engine for Adaptive AI.

Enzyme is an open-source machine learning model designed for flexibility, transparency, and experimentation. Whether you're training on custom datasets, fine-tuning existing logic, or exploring new frontiers in AI, ENZYME provides a fully modifiable Python-based framework.

Ideal for researchers, developers, and curious minds - with ENZYME, you don't just use the model, you make it your own.

---

Gender Classification from Face Images

A machine learning project to classify gender (Male/Female) from face images using deep learning and computer vision. The system includes data collection, preprocessing using YOLOv8n-face, CNN training, prediction, and API deployment.

---

📁 Project Structure

/
├── collect_dataset.py        # Collect raw face dataset via webcam
├── clean_dataset.py          # Clean dataset using YOLOv8n-face (grayscale, cropped faces)
├── trainer.py                # Train CNN model with cleaned dataset
├── predict_img.py            # Predict gender from a single image
├── predict_video.py          # Predict gender from video
├── predict_webcam.py         # Real-time gender prediction via webcam
├── server.py                 # Flask API for prediction
│
├── dataset/                  # Raw images
│   ├── male/
│   └── female/
│
├── cleaned/                  # Preprocessed (cropped + grayscale)
│   ├── male/
│   └── female/

---

🧠 Workflow Overview

1. 📸 Data Collection

Use webcam to collect images categorized into male or female:

python collect_dataset.py

Images are saved under the dataset/ directory.

---

2. 🧼 Dataset Cleaning with YOLOv8n-face

Detect and crop faces using Ultralytics YOLOv8n-face, convert to grayscale, and save:

python clean_dataset.py

• Uses YOLOv8n-face (yolov8n-face.pt)
• Crops faces from images and saves grayscale versions
• Processes multiple categories and images concurrently with multithreading
• Cleaned images are saved in cleaned/

---

3. 🧠 Training the Model

Train a CNN model to classify gender based on the cleaned face dataset:

python trainer.py

Model will be saved as enzyme_model.h5.

---

4. 🔍 Prediction

Predict gender using the trained model:

• From image:

  python predict_img.py

• From video:

  python predict_video.py

• From webcam:

  python predict_webcam.py

---

5. 🌐 API Server

Start the Flask server for remote prediction:

python server.py

Endpoint:

• POST /predict – Accepts image uploads, returns predicted gender.

---

⚙️ Requirements

Install all dependencies:

pip install -r requirements.txt

Libraries used:

• opencv-python
• ultralytics
• tqdm
• flask
• tensorflow or keras

---

📦 Notes

• YOLOv8n-face model file (yolov8n-face.pt) must be available in the same directory or specify the correct path in clean_dataset.py.
• The system uses multithreading to speed up image preprocessing.
• Supports .jpg, .jpeg, .png images.

---

👤 Author Info

Wiji Fiko Teren
🌐 wijifikoteren.streampeg.com
📧 Email: [email protected] / [email protected]
📺 YouTube: @wijifikoteren
☕ Donate: PayPal - paypal.me/wijifikoteren

---

🪪 License

MIT License — use freely for learning and development.