Thanks for your question, sorry for the later response. Yes, you can quantize YOLO V2 with Ristretto. As you said, you first train the network in Caffe, then quantize it using Ristretto. The network has many convolution layers which will get quantized by Ristretto. I have never quantized this particular CNN myself though, so let me know about your results.
Best,
Philipp