deface

In the midst of building a climbing app, I want to use climbing videos of other climbers but I didn't want to show their face without their consent so I tried to look for quick solution with ffmpeg to blur human faces. There is https://github.com/ORB-HD/deface. But why use a ready made software when you can have AI build one for you This project uses media pipe pose detection as a fallback of face detection. The app is hosted on github page as it is completely static. I've only tested in chrome and firefox, with climbing videos (medium range). The detection is jittery so manual editing is supported. The mask can be separately downloaded as a webp, and a pairing ffmpeg command is provided which can be run locally and much faster. Source code at https://github.com/kmcheung12/faceblind

Hey everyone! I've been producing some content for a friend of mine that requires a number of people's faces to be blurred for compliance reasons. In short, *I'm looking for an open source tool that can automatically blur faces in 4k video in rendered output*, (everything is post production no live-streaming). Initially the duration, resolution and format of the video made it possible to do this with the "ai backed" tool in Adobe Premiere. However, now I'm working with 4k footage and the video has shifted to a format with a single face moving too far around the frame for Adobe's tool to be useable or has multiple faces. I'm not a domain expert here, since the last time I did anything with ML was five years ago in college. Even if I could get the model together, writing a pipeline to actually blur 4k video is a mystery to me (although I