Image-Classifier-Flask-App

It is a tool to compare differnt pre-trained Machine Learning Models using Pytorch. Written in Python using Flask.

The API

The API expects a json request as below:

{
    "url": "https://www.tensorflow.org/tutorials/images/classification_files/output_N1loMlbYHeiJ_0.png",
    "model": "squeezenet"
}

where url can be a url for a image file and model can be anything from below array:

["alexnet", "resnet", "squeezenet", "vgg", "densenet", "googlenet", "shufflenet", "mobilenet", "resnext", "wide_resnet", "mnasnet", "efficientnet", "regnet_x", "regnet_y"]

The Response of the API request is:

{
    "clsName1": "bee",
    "clsName2": "bubble",
    "clsName3": "ant, emmet, pismire",
    "img_url": "https://www.tensorflow.org/tutorials/images/classification_files/output_N1loMlbYHeiJ_0.png",
    "nnModel": "squeezenet",
    "percent1": 85.892822265625,
    "percent2": 2.063873529434204,
    "percent3": 1.8604017496109009
}

where clsName indicates the detected object and percent indicates respective confidance percentage.

Here is a screenshot of the API request and it's response in POSTMAN:

Colab Notebook for demo:

Click this link to open Colab Notebook. Run each cell one after another.