Computer Vision Model

Background

Trained a computer vision model from scratch to recognize and classify specific objects in real-time video streams. The project covered data collection, annotation, model training, and iterative refinement based on detection accuracy. It was a deep dive into applied machine learning — understanding not just how to use a model, but how to build, evaluate, and improve one from the ground up.

How it works

1. 01Collected and annotated a custom dataset for the target objects.
2. 02Trained a model and measured detection accuracy on held-out data.
3. 03Iteratively refined the model based on where it failed.

Outcome

• An end-to-end ML pipeline, from data to inference
• Real-time classification on live video
• Hands-on understanding of building and improving a model