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

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.

follow the TransNetV2 inference README and use the NVIDIA DOCKER.

docker build -t transnet -f inference/Dockerfile .

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

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！