AI-Powered Car Feature Detection and Analysis

This project stems from an interesting challenge I observed with the parking attendants in the City of Cape Town—the workers responsible for tracking parking durations and handling payments. Currently, they manually record key vehicle details, including the car's make, color, and number plate.

The process is inefficient: they enter the car make and color manually on a mobile device, while a separate OCR device captures the number plate. With advancements in AI, particularly Large Language Models (LLMs) and computer vision, this task could be streamlined into a single, more efficient system.

With this in mind, I set out to build a demo application that automates the entire process. You can check out my demo app here (This app will work on a mobile phone too).

Note: The application may take up to a minute to load, as it runs on a free backend service that lacks always-on capabilities. If the service is in downtime, the first request will trigger a cold start, requiring a brief wait before the app becomes fully responsive.

Tools used: Javascript, HTML, CSS, Langchain for LLM API calls, Nodejs for backend