FER Benchmarking Platform

Facial Emotion Recognition (FER) Benchmarking Platform is a tool designed to evaluate and compare the performance of different emotion recognition models. It provides a user-friendly interface to run predictions on images using both custom-trained models and state-of-the-art pretrained models, as well as visualize their performance statistics.

Key Features

1. Image Processing & Prediction

Run emotion recognition models on your own images:

• Dual Model Inference: Simultaneously run predictions using:
  • Custom Model: A CNN trained on the FER-2013 dataset.
  • Pretrained Model: A Hugging Face pipeline model (e.g., michellejieli/emotion_text_classifier adapted for images or similar).
• Interactive UI: Upload an image to see a preview and get immediate results.
• Detailed Analysis: View the top predicted emotion, confidence score, and a visual probability distribution for 7 emotions (Angry, Disgust, Fear, Happy, Sad, Surprise, Neutral).

2. Model Statistics & Comparison

Analyze and compare model performance (accessible via API/Future UI):

• Performance Metrics: View confusion matrices, ROC curves, and per-class precision/recall charts.
• Before/After Analysis: Compare the accuracy of the custom model before and after training.
• aggregated Stats: Endpoint access to comprehensive evaluation reports.

---

Installation & Setup

Prerequisites

• Python 3.8+
• Node.js & npm (for the frontend)
• Anaconda (optional, for environment management)

1. Backend Setup (FastAPI)

1. Clone the repository:

   git clone <repository-url>
   cd FERBenchmarkingPlatform

2. Create and activate the environment: Using Conda (recommended):

   conda env create -f env.yml
   conda activate <model-env-name>

   Alternatively, install dependencies manually:

   pip install -r CustomModel/requirements.txt

3. Run the Backend Server: start the FastAPI server from the root directory:

   python server.py

   The API will be available at http://localhost:8000.

2. Frontend Setup (Angular)

1. Navigate to the frontend directory:

   cd frontend

2. Install dependencies:

   npm install

3. Run the Frontend:

   npm start

   The application will be accessible at http://localhost:4200.

---

Usage Guide

1. Start both the Backend and Frontend servers using the commands above.
2. Open your browser to http://localhost:4200.
3. Image Processing Page:
   • Click "Upload Image" to select a face image.
   • Click "Run Models" to process the image.
   • Compare the results (Top Emotion, Confidence, Probability Bars) between the Pretrained and Custom models side-by-side.
4. Comparison Page:
   • Use the navigation bar to switch to the "Comparison" view (Feature coming soon).

Project Structure

• server.py: Main entry point for the FastAPI backend.
• frontend/: Angular source code for the user interface.
• custom_model/: Contains train.py, inference.py, and model artifacts.
• pretrained_model/: Contains the interface for the Hugging Face model.
• evaluation_scripts/: Scripts for evaluating models and generating plots/reports.
• stats_api.py: API endpoints for serving investigation statistics.