Basket Icon Detector

In this project, I've built a Computer Vision model to help a client to detect e-commerces websites.

Problem

• The client had a solution to detect if a website is an e-commerce or not and would like to improve the accuracy of detection.

Task

• Improve the current solution to detect, based on the main website page, if it is an e-commerce or not.

Solution

• The client takes a screenshot of the main page of the website.

• I've built an object detection model using Tensorflow to find basket icons on the screenshot image.

• The model was trained with ~ 3000 images with a diverse set of basket icons.

• The model accuracy was 95.5%

• I've built a web app so the client could test the solution.

Results

• I've created a REST API to serve the model in production and package it all in a Docker container to easily deploy and scale it.

• The client integrates the API in his solution and improves the e-commerces detection.

• Here is the client's feedback about the project:

---

Source code

As this is a private project, the source code is not available.

You can check the tools I've used and try them by yourself or just send me a message so we can work together.

Built With

• Python - Programming Language.

• Tensorflow - An end-to-end open-source platform for machine learning.

• FastAPI - A web framework for building APIs with Python.

• Docker - Industry standard for containers.