ml-image-classification-server

ML-Image-Classification-Server

A server-side application for image classification using K-Nearest Neighbours (KNN) algorithm

• Reads all the image datasets from the /dataset directory
• Allows for image upload for classification
• Accepts PNG, JPG, and JPEG files
• All uploaded images are resized to a standard size and converted to PNG as configured in config.json
• Calculated euclidean distances for input images
• Applies KNN (K-Nearest Neighbours algorithm) to classify images
• The server then returns the result of the classification, i.e., the label of the object in the image

Instructions

Put the image dataset in /dataset directory. Each subdirectory should contain images of a single class or category as shown below:

dataset/
    leopard/
        img01.png
        img02.png
        img03.png
        ...
    laptop/
        img01.png
        img02.png
        ...
    camera/
        img01.png
        img02.png
        ...

To run the server:

    >./ML-Image-Classification-Server

Next, perform requests with new images to get a response from the server classifying them:

curl -F file=@./testimage.png http://127.0.0.1:3055/image

The server will return a response like:

    seems to be a leopard

You can perform tests using the test.sh file:

bash test.sh

Useful commands

To send files over ssh:

scp dataset.tar.gz root@SERVERIP:/root/ML-Image-Classification-Server

To untarr files on the server:

tar -xvzf dataset.tar.gz