Let the driver to open fast forward, a key part of miles deep using the CNN interception of pornogra 1256789

  • hanson
  • November 16, 2017
  • Home Products & Services
  • Comments Off on Let the driver to open fast forward, a key part of miles deep using the CNN interception of pornogra 1256789

Open source | let old driver to fast forward, a key part of Miles Deep using CNN intercept pornographic video Sohu technology selected from the Github machine in the heart of compiling participation: Wu pan in the "illegal" viewing porn video, whether you also for the "long" prelude to regret? A few days ago, GitHub user ryanjay0 open source can be used to identify a specific type of pornographic videos in the scene of the AI project Miles Deep. The algorithm allows you to type a fragment of the type you want to see from the full video and generate a new video that combines these fragments, so that you can no longer worry about the unnecessary fragments. Miles Deep uses a connection with residual (residual connections) the depth of the convolutional neural network (DCNN), can be based on sexual behavior (sexual act) will be a pornographic video every second is classified into 6 types, its accuracy is as high as 95%. It can then be used to automatically edit the video. It can remove all scenes that do not contain sexual contact, or edit out a specific sexual behavior. NSFW model Miles Deep and YAHOO recently released (see "YAHOO open heart machine reported the first pornographic image detection deep learning solutions") using a similar structure, but not the same as the Miles Deep is able to distinguish between sexual behavior and a variety of different specific naked. As far as I know, this is the first publicly available video classification or editing tool. This program can be said that the use of Caffe model for video classification of a general framework, which uses the C++ batching and threading. By replacing the weights, model definitions, and mean file, it can be used immediately to edit other types of video without the need to recompile. Here is an example. Install the Ubuntu installation package (16.04) (Dependencies) sudo apt on install ffmpeg libopenblas-base libhdf5-serial-dev libgoogle-glog-dev sudo apt 14.04 additional dependencies of install libgflags-dev CUDA (recommended) if you need to use GPU, Nvidia GPU and CUDA 8 driver. Highly recommended; can speed up more than 10 times. This can be downloaded directly from the package or CUDNN (NVIDIA) (optional) this is an additional driver from NVIDIA that allows CUDA GPU to support faster. Download here (need to register): Download Miles Deep also want to download this model. Miles-deep and the n相关的主题文章: