ooking for datasets with both structured data and images for new model proof of concept.

Follow the full discussion on Reddit.
I recently worked on a project where I was predicting the likelihood of a car accident at intersections. The dataset contained information about that intersection (number of legs, speed limits, demographical data) as well as GPS coordinates. I fed the GPS coordinates into the Google Maps API to get images of each intersection. So the each row of the dataset contained a list of features, a response variable (number of crashes over 10 years) and the path of the associated image.