I agree - although I think there are three options: 1. Deep learning - therefore any deep learning framework 2. "Classical AI" - use OpenCV for most of it 3.

I agree - although I think there are three options: 1. Deep learning - therefore any deep learning framework 2. "Classical AI" - use OpenCV for most of it 3. APIs as you mentioned above