Hi fellow learners and practitioners 
I’m working with a friend’s company on a classification task where we take an image of a part and classify it (to get the part number). We’ve been shooting images manually so far but now got some research partnership with a university that will enable us to build a device to semi-automatically take the images for us. We’re currently working on the specs for the device and one of the ideas we have is to augment the images with different backgrounds using a green screen background when taking the image. This would enable us to randomly augment the dataset with different backgrounds (which we’ve been doing manually so far). That way we could cut the time it takes to create the dataset tremendously.
Do you have any experience or ideas on the topic?
This is what the current dataset (only about 50 categories) looks like:
Thanks and have a great day! 
Daniel