fiatcode
236e0d2ff2
- Add YuNet face detector as alternative option (built into OpenCV) - Add multi-scale detection (1.0x + 1.5x) to catch faces at different distances - Add NMS to remove duplicate detections from multi-scale - Move frame interval and clustering settings to advanced options - Increase default blur padding from 25% to 40% - Change default frame interval from 30 to 15 - Change default confidence threshold from 0.7 to 0.8 - Add limitations section to README (extreme angles, small faces, motion blur) - Require scikit-learn>=1.3.0 for HDBSCAN support
59 lines
3.2 KiB
Markdown
59 lines
3.2 KiB
Markdown
# PyFaceBlur
|
|
|
|
An interactive command-line tool that automatically detects, clusters, and blurs faces in videos. It guides you through a simple step-by-step process to extract frames, group people by facial identity, select who you want to blur, and re-encode the video.
|
|
|
|
## Features
|
|
|
|
- **Interactive CLI:** Built with `rich` and `questionary` for a clean, prompt-based UX including file path auto-completion.
|
|
- **Accurate Face Recognition:** Uses [UniFace](https://github.com/yakhyo/uniface) (RetinaFace detection + ArcFace 512-dim neural embeddings via ONNX Runtime) to accurately re-identify the same person across a video.
|
|
- **DBSCAN Clustering:** Automatically groups identical faces into "clusters" using Cosine similarity.
|
|
- **Hardware-Accelerated Encoding:** Automatically detects and leverages GPU encoders like `av1_vaapi`, `hevc_vaapi`, `h264_vaapi`, `h264_nvenc`, and more via FFmpeg.
|
|
- **Visual Face Selection:** Extracts one high-quality thumbnail per detected person and opens your system's file explorer so you can easily check boxes for who to blur.
|
|
- **Multiple Blur Styles:** Choose from Gaussian, Pixelate, Blackout, Elliptical, or Median blur methods.
|
|
- **Smooth Interpolation:** Bounding boxes are linearly interpolated between sampled keyframes and held static when faces exit/enter, ensuring smooth blurring without split-second exposures.
|
|
|
|
## Requirements
|
|
|
|
- Python 3.11+
|
|
- [uv](https://docs.astral.sh/uv/) for fast dependency management
|
|
- `ffmpeg` installed and available in your system `$PATH` (for frame extraction and re-encoding)
|
|
|
|
## Setup
|
|
|
|
```bash
|
|
# Clone the repository and navigate to the project directory
|
|
cd py-faceblur
|
|
|
|
# Sync dependencies using uv
|
|
uv sync
|
|
```
|
|
|
|
## Usage
|
|
|
|
Run the interactive wizard:
|
|
|
|
```bash
|
|
uv run pyfaceblur
|
|
```
|
|
|
|
### The Pipeline
|
|
|
|
1. **Input:** You provide the path to your video and the frame sampling interval (e.g., sample every 30th frame).
|
|
2. **Processing:** The app uses FFmpeg to extract frames, runs RetinaFace to find all faces, and generates ArcFace embeddings.
|
|
3. **Clustering:** DBSCAN groups the embeddings to identify unique individuals.
|
|
4. **Selection:** The app saves a thumbnail of each person to a temporary folder, opens it, and asks you to select which people to blur using interactive checkboxes.
|
|
5. **Encoding:** The app finds the best available video encoder on your system, applies the chosen blur method to the selected faces, interpolates their movement, and generates a new `*_blurred.mp4` video.
|
|
|
|
## Advanced / Legacy CLI
|
|
|
|
The original proof-of-concept command-line interface is also still available for purely extracting and debugging the clustering outputs into an output folder.
|
|
|
|
```bash
|
|
uv run pyfaceblur-legacy detect --video input.mp4 --output ./output --interval 30 --confidence 0.7
|
|
```
|
|
|
|
## Limitations
|
|
|
|
- **Extreme face angles:** Faces viewed from extreme angles (e.g., strong profile views, looking up/down) may not be detected or may be clustered as separate identities. For best results, use videos where faces are mostly front-facing or at moderate angles.
|
|
- **Small/distant faces:** Very small faces (below 50 pixels) may not be reliably detected or produce accurate embeddings for clustering.
|
|
- **Rapid motion blur:** Fast head movements causing motion blur can affect detection accuracy.
|