llanopoplar

3 ms·

1. 1.▲
   by llanopoplar 8y ago

   It loads dnn models via OpenCV api(caffe, darknet, tensorflow, etc) and feed video streams to various detectors. Support run time add/delete detectors. Just for fun.
2. 2.▲
   Show HN: A play case: add/delete object detector during transcode(github.com)

   2 points by llanopoplar 8y ago|1 comments
3. 3.▲
   by llanopoplar 8y ago

   Besides, 1. it provides a way can compose different audio/video component. For instance, video decode -> post-video decode filter -> video mix -> pre-encode fileter -> video encoder -> muxer. 2. can easily write new plu
4. 4.▲
   by llanopoplar 8y ago

   Not for editing. A library can do dynamic change during transcoding or live broadcast, add new inputs/output, change output settings. It is more like a Video Conference backend, but keep FFmpeg's libav* capacity.
5. 5.▲
   by llanopoplar 8y ago

   The code shipped with an application 'Interactive Live' which builds on FFdynamic. It shows how concise code could be, when we can make audio/video process in a compositional way, to do dynamically add input streams, add new
6. 6.▲
   Show HN: FFdynamic – Extend FFmpeg with A/V process composition, runtime control(github.com)

   42 points by llanopoplar 8y ago|5 comments