Computer vision @ Infosys

Facial Recognition & Recommendation

A computer-vision facial recognition and recommendation system built with Python and OpenCV. Preprocessing plus deep-learning feature-extraction pipelines exposed via REST APIs to a web UI.

OpenCVPythonDeep LearningREST APIs

Role: CV intern @ Infosys
Timeline: 2023

View on GitHub ↗

Facial recognition UI — 'It doesn't guess who you are. It measures it.' with 96% match, embedding distance 0.042, 38ms inference and recommendation cards — Live match — 96%, 38ms

detect → embed → match pipeline with capture/detect/embed/match/respond CLI, precision 0.94, recall 0.92, f1 0.93, 11ms GPU inference — Pipeline — precision 0.94

Problem

Build a facial-recognition system that also recommends related profiles/items, exposed through a simple web UI.

Approach

01Preprocessed images (alignment, normalisation) with OpenCV.
02Extracted embeddings with a deep-learning feature extractor.
03Matched embeddings via nearest-neighbour search for recognition + recommendation.
04Exposed everything as REST APIs consumed by a lightweight web UI.

Stack

OpenCV
Python
Deep Learning
REST APIs

Outcomes

→End-to-end pipeline from raw image to ranked recommendations.
→Reusable API contract handed off to the product team.

View on GitHub ↗