Is it hard to make a machine learning model?

Custom Vision or CV.AI might be the right free service to start building your own state-of-the-art computer vision models that fit perfectly with your unique need and dataset.

How many steps to train a ML model?

  1. Get the dataset, I suggest you to use 4k stogram to download images from specific hashtags from Instagram
  2. Upload the images to CV.AI, grouped by tag/class
  3. Train them in less then one minute
  4. Evaluate it with simple new images
  5. Use the trained TensorFlow Frozen Model with
    1. Python tutorial
    2. Javascript tutorial:
      1. convert the Frozen Model with tensorflow tfjs-converter;
      2. use the commands at the end of this notebook.
      3. use this ModelLoader class and modify it with your model url and also a classes file like this one classes.js converted from the labels.txt downloaded from

Is it really easy?

Not entirely in fact after the end of this exam period I’m going to build a simple website that is going to do all the previous steps for you, if I’ll find some spare time after doing Augmented Reality all the day.