"CloudCV ModelHub: A Collaborative Model Registry with Automated Benchmarking" #76
Description
Project Idea: CloudCV ModelHub
Description: A web platform for CloudCV where users can upload, browse, and download pre-trained computer vision models, with automated benchmarking on standard datasets. Results are displayed in an interactive AngularJS dashboard, and models are containerized with Docker for reproducibility.
Features:
- Model registry with Django backend and AWS S3 storage.
- Automated benchmarking using Python and Docker, executed on AWS.
- AngularJS frontend for browsing and visualizing results.
- Reproducibility via downloadable Docker images and collaboration via comments.
Impact: Enhances CloudCV’s ecosystem by providing a collaborative hub for sharing and comparing vision models, making AI research more reproducible and accessible.
Skills: Python, Django, Docker, AngularJS, AWS, computer vision, deep learning.
Mentor Notes: Could integrate with EvalAI or support specific model formats (e.g., ONNX).