Fast.ai for training an object detection model for a specific type of object/activity?

I’m interested in training an object detection model for a specific type of object and/or activity which is not easily detected using any of the pre-trained models I’ve found. I am hoping to build a model that will aid in detecting security threats captured on video by security cameras.

I think this can easily be done using other ML frameworks – for example, see this article for an example of someone training a model for detecting a Rubik’s cube.

I’d like to give this a whirl using fast.ai if it’s possible, as I assume that the fast.ai approach will be simplest for someone like me who is still an ML beginner. Can anyone please advise?

Thanks in advance for any suggestions or comments.

1 Like

It looks like this is underway, another user has asked a similar question. The development notebook for this is a bit daunting. :frowning:

A response to the above question lists a project where a user has made good progress with fast.ai using an SSD, and based just on the number of lines of code it looks to be an easier solution (from my naive perspective anyway).

This may be relevant as well: Training an Image Classification Convolutional Neural Net to Detect Plant Disease Using fast.ai