Movie Finder

The Challenge

Users often see a movie scene and can't identify the film. Existing solutions required text descriptions, which are imprecise and frustrating. The challenge was building an AI system that could identify movies from visual input — a single frame or short clip — with high accuracy and speed.

Our Solution

We built a custom ML pipeline using CoreML and Vision Framework on iOS, backed by a TensorFlow model trained on millions of movie frames. The app processes images locally for speed, then queries our cloud API for final identification. We used GCP for scalable model serving and built a proprietary frame-matching algorithm that handles different aspect ratios, color grading, and compression artifacts.

Key Features

Results