Model Evaluation Website

Introduction

Welcome to our Model Evaluation Platform! Designed with cutting-edge capabilities, our platform allows users to thoroughly assess their machine learning models across several critical attributes. These include Correctness, Robustness, Fairness, and Transferability—each supported by a diverse range of evaluation perspectives and metrics.

Our platform provides flexibility in customizing evaluation tasks, datasets, metrics, and model parameters, giving you full control over how models are assessed. Users can also visualize evaluation results, gaining a clearer, intuitive understanding of model performance. This empowers you to optimize and improve your models more effectively.

Currently, the platform specializes in key computer vision tasks, including Image Classification, Object Detection, and Image Segmentation. We are committed to continually evolving the platform's features, ensuring it stays in sync with advancements in deep learning technologies and application needs, and providing our users with increasingly powerful evaluation tools.

Properties

Our platform supports model evaluation across the following key attributes:

• Correctness
• Robustness
• Fairness
• Transferability

---

Frontend and Backend Setup

Backend Setup

1. Create and activate a Conda environment by running the following command:

   conda create --name model-evaluation python=3.9
   conda activate model-evaluation

2. Install the required dependencies:

   pip install -r requirements.txt

3. To start the backend, run the start.sh script:

   • For development environment, add the dev flag:

     bash ./src/start.sh dev

   • For production environment, run without the dev flag:

     bash ./src/start.sh

   The script will handle migrations and then either start the Django development server (runserver) or use Gunicorn for production (gunicorn).

Frontend Setup

1. Install nvm (Node Version Manager) if it's not already installed.

2. Install Node.js version 16.20.2:

   nvm install 16.20.2

3. Switch to Node.js version 16.20.2:

   nvm use v16

4. Install Yarn package manager for the selected Node.js version:

   npm install -g yarn

5. Install the necessary frontend dependencies:

   yarn install

6. To build the frontend project:

   yarn build

---

Contact

This project is developed by VIPA Lab from Zhejiang University.

---