Background

Nowadays cameras are distributed everywhere, from the streets of Beijing, to New York, from the intersections of Santa Clara to the parks we talk a stroll or the street lamps that we drive under.

Police cars have them, while casinos also have them.  Someone predicts that there will be over billion cameras by 2021. In a word, everywhere we go, our activities are being constantly recorded. It's almost impossible for human being to watch and analyze these tremendous video generated by those security cameras.

Solution

The deep learning application scene becomes very diverse in Smart cities, not only including detection, segmentation, but also classification and tracking. In many cases, we may need to combine all of these processing in a single application. Moreover, accurate result of these operations is super important for helping higher level applications to generate very useful outputs.