---
license: apache-2.0
library_name: huggingface_hub
pipeline_tag: image-segmentation
tags:
- model_hub_mixin
- pytorch_model_hub_mixin
---

# PPEDCRF: Privacy-Preserving Enhanced Dynamic CRF

PPEDCRF is a framework for location-privacy protection in dashcam videos. It identifies and tracks location-sensitive background regions using a dynamic Conditional Random Field (CRF) and applies calibrated perturbations to these regions while preserving foreground detection and segmentation utility. This repository contains the sensitive-region predictor (**SensNet**) component of the framework.

- **Paper:** [PPEDCRF: Privacy-Preserving Enhanced Dynamic CRF for Location-Privacy Protection for Sequence Videos with Minimal Detection Degradation](https://huggingface.co/papers/2603.01593)
- **Code:** [GitHub Repository](https://github.com/mabo1215/PPEDCRF)

## Usage

This model is intended to be used with the official [research codebase](https://github.com/mabo1215/PPEDCRF). 

### Installation

```bash
git clone https://github.com/mabo1215/PPEDCRF.git
cd PPEDCRF
pip install -r src/requirements.txt
```

### Video Protection

To use this pre-trained checkpoint to protect video clips:

```bash
python src/main.py --config src/config/config.yaml protect --checkpoint mabo1215/ppedcrf-sensnet
```

## Citation

```bibtex
@article{ma2024ppedcrf,
  title={PPEDCRF: Privacy-Preserving Enhanced Dynamic CRF for Location-Privacy Protection for Sequence Videos with Minimal Detection Degradation},
  author={Ma, Bo and Wu, Jinsong and Yan, Weiqi and Shi, Catherine and Nguyen, Minh},
  journal={arXiv preprint arXiv:2603.01593},
  year={2024}
}
```