How do I train the Azure Video Indexer on detected objects?

Question

My organization has years of high quality footage of bald eagle and peregrine falcon nest life that we want to catalog. I am trialing the Azure Video Indexer to see if it can help us build a library. Unfortunately, almost none of my objects - eggs, eggshells, eaglets, and so on - are supported objects. Is it possible to train the detector to recognize unsupported objects?

In the example above, the indexer identified bald eagle eggs and a bald eagle's head as 'sports ball', which is extremely amusing (not 'golf ball' or 'baseball'?) but not very helpful when it comes to organizing content for potential use.

Answer 1

@Amy Ries I think you could try to use the Azure custom vision service for your use case. Custom vision allows you to upload images of your domain to train a custom model to identify images and classify them as per the tags that were used during training. I will recommend taking the classify images custom vision learn course with the custom vision portal to check if this helps your use case. Thanks!!

If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.