DFACE.app

DFACE.app is a web application that performs private, automatic face redaction in a web browser using YOLOv5 computer vision models and Tensorflow.js. Made by VFRAME.io.

Environment Setup

Install NVM from https://github.com/nvm-sh/nvm

nvm use node
nvm install 14.17.1
nvm use 14.17.1

Install Yarn

npm install -g yarn

Install dependencies

yarn install

Disable Next.js telemetry:

npx next telemetry disable

Development

Start a development server on http://localhost:3000/

yarn dev

Models should be placed in public/assets/models/ in folders named identically to the names in modelzoo.yaml. The application will look for the file model.json when loading a model. The active models can be rebuilt from the YAML by running:

yarn modelzoo

Emoji

Let's admit it, emoji are probably the most important part of this application. Here is how you can generate a test document of emoji so you can load more (using twemoji):

cd node_modules/twemoji-emojis/vendor/svg
echo "<style>div img { width: 48px; height: 48px; display: block; } div { display: inline-block; font-size: 12px; }</style>" >> emoji.html
for i in *.svg; do echo "<div><img src="$i">$i</div>" >> emoji.html; done
python3 -m http.server 8000

Navigate to http://localhost:8000/emoji.html
Emoji are listed in components/common/Emoji.js, follow the pattern to add new ones
Emoji are grouped by classes. The list of classes should correspond to the emojiOptions in constants/index.js

Deployment

This command will build the static site to the out/ directory, suitable for deployment.

yarn build

Credits

DFACE uses the YOLOV5 object detection architecture by Glenn Jocher based on the original YOLO by Joseph Redmon

Acknowledgments

VFRAME gratefully acknowledges support from NLnet for the continued development of DFACE:

VFRAME received support from the NLNet Foundation and Next Generation Internet (NGI0) supported research and development of face blurring and biometric redaction tools during 2019 - 2021. Funding was provided through the NGI0 Privacy Enhancing Technologies Fund, a fund established by NLnet with financial support from the European Commission’s Next Generation Internet program.

Licenses

DFACE web application MIT
Next.js: MIT
Tensorflow.js: Apache 2.0

tttimur / dface