Product-Recognition-on-Store-Shelves

What is this repository for?

• This repository implements two object detection algorithms based on SIFT

Package equirements

Python version 3.9
opencv-contrib version 4.4.0.44 jupyter version 4.6
numpy version 1.22.1

Overview

A project about object detection techniques based on computer vision. The algorithm can be deployed to identify query images contained inside a scene, the provided example shows an use case of product recognition on store shelves. In order to work properly, the algorithm needs two inputs:

• A scene image, containing the query(ies) to be found.
• A query image, the image to be recognised.

Functioning

The project deploys two different algorithms in order to achieve object detection:

Mean-Shift clustering

This version uses mean-shift clustering to partition the detected keypoints into different clusters, then, for each cluster, matching is computed separately.

Generalized Hough Transform - Star Model

This version uses GHT along with SIFT features, each matched keypoint casts a vote for where the barycenter is, then the resulting grid is thresholded to find objects.

Who do I talk to?

• Luca Rispoli - [email protected]
• Marco Ghaly - [email protected]