# Real-Time Person Anonymizer

Background-subtraction person anonymizer in pure OpenCV — no ML model required.

## Overview

`Final.py` detects people in a live webcam stream using classical background subtraction and change them with a mosaic or Gaussian blur in real time. The system learns a static background over a 10-second warmup, then masks subsequent motion through a five-stage pipeline (foreground diff, morphology cleanup, contour merging, feathered anonymization). It runs on any laptop CPU with only OpenCV and NumPy.

## Features

- Real-time person detection without a pre-trained model
- Two anonymization modes: irreversible mosaic and Gaussian blur
- Interactive camera setup with exposure / white-balance / focus sliders
- Live threshold tuning during the main loop
- "Too close" protection (disables blurring when the person mask exceeds 55 % of the frame)
- XVID-encoded `.avi` recording of the anonymized stream

## Requirements

- Python 3.x and a webcam
- `pip install opencv-python numpy`

## Quick start

```
python Final.py --camera 1
```

## CLI arguments

| Flag | Default | Description |
|---|---|---|
| `--camera N` | 1 | Camera index |
| `--output FILE` | timestamped `.avi` | Output filename |
| `--width N` | 640 | Capture width |
| `--height N` | 480 | Capture height |

## Keybindings — setup screen

| Key | Action |
|---|---|
| `[` / `]` | Previous / next camera |
| Mouse | Buttons and sliders |
| `SPACE` | Proceed to warmup |
| `q` | Quit |

## Keybindings — main loop

| Key | Action |
|---|---|
| `q` | Quit |
| `b` | Re-capture background |
| `+` / `-` | Decrease / increase BGR threshold |
| `[` / `]` | Decrease / increase HSV threshold |
| `m` | Toggle mosaic ↔ Gaussian |
| `n` / `N` | Decrease / increase mosaic block size |
| `1` / `2` / `3` | Overlay / anonymized (records) / debug |
| `p` | Pause |

## Project structure

- `Final.py` — main application (originally `Modified_Blur_V3_1_1.py`)
- `Modified_Blur*.py`, `person_blur.py` — earlier versions, kept for reference
- `modified_blur_pipeline_workflow.svg` — five-stage pipeline diagram
- `PROJECT_DOCUMENT.md` — full design document

## Known limitations

- Requires a static camera and a person-free warmup scene
- Lighting drift invalidates the background (press `b` to recapture)
- Slow movers are absorbed into the background
- Fingertip dropouts may persist

See `PROJECT_DOCUMENT.md` for tuning guidance.