Detecting rotated, scaled, re-colored images using OpenCV

Solution for Detecting rotated, scaled, re-colored images using OpenCV
is Given Below:

I’ve recently come across a new type of captcha and am just curious about how someone would go about writing code in Python and OpenCV to solve it.

I’m a beginner in the field of image processing and am reading a lot about it online. I’ve written some basic code to solve some simple text-based captchas, but this one has me stumped. If you notice, the objects have a lot of properties modified i.e. they’ve been scaled, rotated, made transparent, re-colored (change of hue, saturation, brightness etc.) and possibly more.

This seemed similar to solving ReCaptcha, so I read up about that. It seems the most common way to solve something like this is to use a neural network and some form of learning algorithm. I’m interested where I could learn more about this.

However, what if I don’t have enough training images to actually make a learning algorithm possible? If I have only 100 or 200 images, that wouldn’t be enough. What can be done in this situation?

I am a beginner here, so please forgive my lack of knowledge here. Also request you to answer keeping this in mind.

Example Captcha