Faster R C N N
detects objects based on image classification model, in our case Resnet50 that is trained to divide images into 1000 categories.
It's too slow for real time use because it needs to run the image classification model on all the parts of the image separately to determine if they contain any object.