HotShot-AI

Inspiration Many people take way too many photos on their vacation trips, and as a result, have many average-looking photos cluttering their albums. We created HotShot-AI to help amateur photographers automatically filter out their best photos from the rest using AI.

What it does HotShot-AI allows users to upload their vacation photos and uses a combination of image quality scoring (sharpness, lighting) and scene detection to analyze each image. We then send this analysis to an LLM (ChatGPT) to generate human-like feedback and rank the images by "vibe"—highlighting the best shots from the batch.

How we built it Frontend: React.js Backend: Node.js, Express, Supabase, PostgreSQL, AWS S3 buckets ML Models: Places365, BRISQUE, OpenCV, OpenAI

Challenges we ran into Rate limiting and quota issues with the OpenAI API Parsing and organizing predictions from the image classifier Designing an intuitive UI under tight time constraints Managing user sessions and photo grouping cleanly

Accomplishments that we're proud of End-to-end functional product: upload, analyze, and rank photos LLM integration for meaningful photo feedback Sleek, responsive UI with album previews and per-image scoring Collaboration across frontend and backend in a short time

What we learned How to use image classification and quality scoring to drive LLM prompts Working with S3 buckets for file handling Clean component-based UI design in React Practical integration of AI into user-facing applications

What's next for HotShot-AI Let users create custom albums and share top photos Use more advanced computer vision models (e.g. CLIP, aesthetic scoring models)