Skip to content

Latest commit

 

History

History
37 lines (27 loc) · 1008 Bytes

README.md

File metadata and controls

37 lines (27 loc) · 1008 Bytes

course-video-to-doc

Extract ppt of online course video with intelligent split shot algorithm

用智能分镜头算法提取网上课程视频的ppt

step-1 choose a good scence clip algorithm

git clone https://github.com/soCzech/TransNetV2.git

TransNet V2: Shot Boundary Detection Neural Network

download the network weights. Does file transnetv2-weights/saved_model.pb exist? It's on git-lfs so maybe git clone did not download it.

step-2 build the environment

follow the TransNetV2 inference README and use the NVIDIA DOCKER.

docker build -t transnet -f inference/Dockerfile .

step-3 video to doc

pip install img2pdf moviepy

replace the inference's code transnetv2.py with our code inference/transnetv2.py

python transnetv2.py ../test.mp4 --threshold 0.5 --pdf --frame-type triple # to get a pdf

TODO

  1. suport set a box area in the video
  2. faster pipeline
  3. better scence clip algorithm

欢迎大家提出修改意见！