Image Classification Program

Image classifier for Philips products.

---

Table of Contents

• Installation of Docker Image
• Clone
• Setup
• Presentation
• Colab
• Dataset Description
• Training
• Validation
• Contact

---

Installation

• After cloning the repository, please remove README.md file in model/val-images
• Before building Docker image, please add test images into model/val-images folder.
• After cloning, please apply Docker command below in the /model/ directory of the project to prevent path errors:

Clone

$ git clone https://github.com/BxCvd1LZVDCvW74I/model.git

Setup

• Tested on Docker Toolbox Version: Docker version 19.03.1, build 74b1e89e8a on Windows 10 Home edition

$ docker build -t philips-case -f Dockerfile .

Presentation

• For this case, I prepared a presentation about my approach to establish an image classification model, including dataset description and training results.

• Link: Presentation

Colab

• In this Google Colab notebook, I reproduced the results due to possible errors from Docker Toolbox for Windows 10 Home edition:
• Please run the notebook until cloning the repository.
• After cloning this repository is done, please upload test images to model/val-images folder and run the other cells to reproduce results.
• Link: Colab Notebook

Training

Validation

Dataset

Image dataset created by using product review videos from Youtube.

• Wake up light : 13.311 images
• Tootbrush : 13.999 images
• Shaver : 14.519 images
• Smart Baby Bottle : 10.503 images

Contact

I'll be checking my e-mail through the competition. Contact information:

• E-mail: [email protected]
• Website at mburakbozbey.github.io
• Linkedin at @mburakbozbey