I have a question regarding Visual AI and the insights that can be derived through the Understand section in DataRobot.
I noticed that the Image Embeddings and Activation Maps can only be obtained for the validation set in the UI. Is there a way to extract them for the whole training data, as well as the scoring later on? Maybe through the API?
While I didn't try this, from reading the DataRobot Python API VisualAI functions, it looks like it is possible. Please find the link to the VisualAI Python manipulation functions
Here is the answer from my colleague, which confirm that you should be able to do it in our Python API
You can extract activation maps for predict data via prediction explanations in UI or Public API - if image column will be among the top explanations it will provide the activation map for it