Hello!
Keeping the data collection and acquisition apart, we can go with the design of classification task ahead and complete in reasonable time of 5 to 7 days.
We can do so using the widely popular AI or neural networks, CNN. I will vote for Python with third party libraries, so it should be cross compatible across platforms.
That will be the algorithm that will run whenever you make a call to an api over the web or whichever your service is.
Regarding your project:
I understand that you need a custom image classification system algorithm for your own set of images. I can give you a codebase, maybe share it on github project for you to visit and download and see how it works.
I shall also try first to use transfer learning such as the available small networks like VGG or AlexNet. Will that be enough?
Development platform
Linux Ubuntu, Python, TensorFlow with Scikit-learn, scipy or opencv if needed for image pre-processing.
Let me know what's the current status and other things. Thank you! Have a great day!
About me:
I'm a research assistant in computer vision and machine learning lab in my university.