Some people might be interested in understanding more about the successful 2nd place method which uses Detection Networks which has strong application potential. Reading the trio of R-CNN papers can be quite heavy going, even if you use the tutorials, particularly for convnet specialists because they also draw on prior detection work a lot. So you might want to start with CS231n lecture 8 [here](https://www.youtube.com/watch?v=_GfPYLNQank&t=513s).