title	emoji	colorFrom	colorTo	sdk	sdk_version	app_file	pinned	license	short_description
Engagement Level Analysis For Single Person Video Clips	📊	red	blue	gradio	6.2.0	app.py	false	apache-2.0	AI-powered human engagement analysis using ViT

Engagement Analysis System

An optimized deep learning application for analyzing and visualizing human engagement levels in video content using Vision Transformers (ViT) and GRU architectures.

🚀 Features

• Adversarial ViT Backbone: High-accuracy facial feature extraction
• Temporal Analysis: GRU integration for consistent engagement tracking over time
• Real-time Visualization: Dynamic bounding boxes with color-coded engagement levels
• Performance Optimized: Batch processing and frame sampling for faster inference

---

🛠️ Installation

1. Requirements

Ensure you have Python 3.10+ installed.
Install the necessary dependencies using:

pip install -r requirements.txt

2. File Structure

The system expects the following data structure (based on the space files):

├── best_model.pth
├── face_detection_yunet.onnx
└── test_samples/
    └── Class_X_Example.mp4

📊 Engagement Levels

The system classifies engagement into four categories based on the calculated score 𝐿:

Level	Range	Visualization Color
Very High	$L \geq 2.5$	Green
High	$1.5 \leq L < 2.5$	Yellow / Cyan
Low	$0.5 \leq L < 1.5$	Orange
Very Low	$L < 0.5$	Red

💻 Usage

Running the UI

Execute the main script to launch the Gradio web interface:

python app.py

Advanced Settings

1. Batch Size: Balance speed vs. VRAM usage (Default: 12)

2. Smoothness: Control the temporal averaging filter (Default: 5)

3. Analysis FPS: Adjust the density of inference frames (Default: 5)

📜 Requirements List

gradio
opencv-python-headless
torch
torchvision
timm
albumentations
numpy