CNN大规模视频处理--NoScope: 1000x Faster Deep Learning Queries over Video

http://dawn.cs.stanford.edu/2017/06/22/noscope/
https://arxiv.org/abs/1703.02529

YOLOv2在视频检测中的效果比较好，但是一个GPU也只能达到每秒几十帧的处理速度。对于上百路视频怎么使用YOLOv2来完成检测和检索的任务了？总不能每一路视频都配置个 GPU吧。这里主要的思路还是先进行运动检测，看看当前帧有没有运动物体models that detect differences (to exploit temporal locality locality)，然后再对每个相机训练一个小的 CNN 模型来完成检测任务。models that are specialized to a given feed and object (to exploit scene-specific locality) .
NoScope’s specialized models can run at over 15,000 frames per second compared to YOLOv2’s 80 frames per second

If the difference detector is confident that nothing has changed, NoScope drops the frame; otherwise, if the specialized model is confident in its label, NoScope outputs the label. And, for particularly tricky frames, NoScope can always fall back to the full CNN.