What is best strategy to recognize paintings?

I have a dataset with 50k photos of paintings. All with a unique ID. I now want to create a phone app (javascript) that can find the unique ID when the user is taking a photo of a painting with the mobile camera.
What might be the best strategy for acomplishing this? Single shot learning? TS Lite?


Welcome to the Tensorflow Forum!

You can use classic computer vision approaches like calculating the feature descriptors for every image and store them in the database along withe their unique id’s. For any new image captured through the camera, the feature descriptor can be calculated and the unique id of the nearest feature descriptor to the test image can be fetched.

Thank you!