There’s another article by Amitabha Banerjee regarding image recognition via Open AI clip. That is a new deep learning algorithm that is able to do object recognition without the need to first train a neural network with a bunch of images yourself. This is done via CLIP (Contrastive Language–Image Pre-training). CLIP connects to a vast library of scanned images on the net instead of depending on a locally trained neural net.
Be aware of the fact that this is advanced stuff and my not be comprehensible to less advanced users. You will need some coding experience. But this could lead to interesting options with Vector in the future via OSKR.