galdric

machine learning server, for image classification

• Reads all the datasets in the folder /dataset
• Runs a server that allows to upload images to classify
• Accepts PNG, JPG, JPEG
• Each image is resized to the same size and converted to PNG type, configured in the config.json
• For the input images, calculates the euclidean distances
• Applyies KNN (K-Nearest Neighbours algorithm) to classify the images
• Server returns the classification result, that is the label of the object in the image

Instructions

Put dataset in /dataset directory with subdirectories, where each subdirectory contains images of one element.

For example:

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

So, we have each image and to which element category is (the name of subdirectory).

Then, run the server:

    >./galdric

Now, just need to perform petitions with new images, to get the response from the server classifying them:

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

And the server will return:

    seems to be a leopard

Can perform some tests with the test.sh file:

    bash test.sh

---

Useful commands

send file over ssh:

scp dataset.tar.gz root@SERVERIP:/root/galdric

on the server, untar file:

tar -xvzf dataset.tar.gz