[D] Detecting arbitrary objects in images
All object detection approaches I know are trained with some data sets to detect only specific classes from the respective data set (also e.g. something like YOLO9000 which can detect 9000 classes). I want to have some more general approach, that can detect arbitrary objects (of any class, returning the containing bounding box) in images.
I did not found anything here, can anyone recommend a related paper or website?