There is the RasterObjectDetector and corresponding RasterObjectDetectionModelTrainer.

However, what little I've seen (out of the box) was a bit hit and miss in terms of the detection, so if this is important to you, I'd recommend looking at something like AWS Rekognition in stead. It's fairly simple to upload images to S3 with FME, then trigger AWS Rekognition (e.g. using a small Lambda script). It will return a json document with bounding boxes and descriptions for all the found objects, which makes it easy to process in FME.

Personally, I've been pretty much blown away by the quality of the results from AWS Rekognition.

Thanks very much for the reply David I will look into the AWS software!