We are getting there, with as usual some trade-off between quality and speed.

http://cs231n.stanford.edu/slides/winter1516_lecture8.pdf shows some benchmarks (mAP = Mean Average Precision, higher is better; FPS = frame per second):

[![enter image description here][1]][1]

[![enter image description here][2]][2]


  [1]: https://i.sstatic.net/AfHt2.png
  [2]: https://i.sstatic.net/B5Qb9.png