AravinthKumar24/Real_Time_Face_Segmentation_for_Movie_Cast_Identification

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

🎬 Real-Time Face Segmentation for Movie Cast Identification

📌 Project Overview

This project focuses on detecting and segmenting human faces in movie scenes using deep learning techniques. The goal is to enable a "Pause-and-Identify" feature in streaming platforms, where users can instantly view actor details.

🚀 Features

• —Face segmentation using U-Net architecture
• —MobileNetV2 encoder (transfer learning)
• —Real-time inference using Streamlit
• —Image upload and webcam support
• —Performance dashboard with metrics
• —Downloadable logs and results

🧠 Problem Statement

Streaming platforms need a system that can:

• —Detect faces in complex movie scenes
• —Segment faces accurately
• —Identify actors in real-time

💼 Business Use Cases

• —🎥 Pause & Identify actors
• —🎯 Personalized recommendations
• —🛡 Content moderation
• —📢 Targeted advertising

🗂 Dataset

• —Movie scene images
• —Corresponding binary face masks
• —Preprocessing:
• —Resize to 256x256
• —Data augmentation (flip, rotate, brightness)

🏗 Model Architecture

• —U-Net architecture
• —MobileNetV2 as encoder
• —Skip connections for better localization
• —Custom Dice Loss function

⚙️ Training Details

• —Optimizer: Adam
• —Loss: Dice Loss + Binary Crossentropy
• —Metrics: Dice Coefficient, IoU
• —Techniques:
• —Early Stopping
• —Model Checkpoint
• —Learning Rate Scheduling

📊 Evaluation Metrics

🔍 Results

• —High accuracy in face segmentation
• —Works well in crowded scenes
• —Real-time performance achieved

🖥 Streamlit App

Features:

• —Upload images
• —Real-time webcam detection
• —Visualization of masks
• —Metrics dashboard

Run App:

streamlit run app.py

📦 Installation

git clone <repo_url>
cd project
pip install -r requirements.txt

📁 Project Structure

├── data/
├── models/
├── notebooks/
├── app.py
├── requirements.txt
├── README.md

📌 Future Improvements

• —Add face recognition (actor identification)
• —Improve inference speed
• —Deploy on cloud (AWS / HuggingFace)

🧑‍💻 Tech Stack

• —Python
• —OpenCV
• —TensorFlow / Keras
• —Streamlit
• —NumPy, Pandas, Matplotlib

🙌 Acknowledgements

• —Open-source community
• —TensorFlow & Streamlit contributors