wtf is egoblur?

EgoBlur is a privacy-protection tool that automatically detects and blurs out faces and license plates from images and videos. You point it at a photo or video file, run the command, and it outputs a version where all the sensitive information has been pixelated. This is useful for anyone who needs to share video footage or images publicly but wants to protect people's privacy and identity. The tool works by using AI models trained to recognize faces and license plates. When you run the command, it scans through your image or video frame-by-frame, finds anything that looks like a face or license plate, and blurs that region. You can adjust how aggressive the blurring is and which types of things to blur, just faces, just license plates, or both. The README mentions two versions of the models: Gen1 (for older Aria camera devices) and Gen2 (for newer ones), so you need to use the right version that matches your camera hardware. People would use this if they're working with video from body cameras, dashcams, or security footage and need to anonymize it before sharing or publishing. For example, a researcher recording people in public might need to blur faces before releasing the data. A news organization processing dashcam footage could use it to protect bystanders. Or someone publishing a walkthrough video might want to blur license plates of parked cars. The tool runs on Python and can handle common formats like MP4, PNG, and JPEG files. One notable aspect is that the tool is built by Meta research and is open-source, so you can download and use it for free. It supports GPU acceleration through CUDA if you have a compatible graphics card, which speeds things up considerably. There's also a specialized version that works directly with VRS files (a Meta Aria video format) without needing to convert to MP4 first, which preserves other sensor data like motion tracking.