All Projects

Blurify.AI.

The Streamlit app built on top of Face Blurring — upload an image, pick YOLOv8 or Haar Cascade, get the faces blurred in-browser.

Computer VisionArchivedMITStreamlitYOLOv8n · Haar CascadeDocker
2
Model choices
25×25
Blur kernel
Python 3.7
Base image
Face Blurring
Built on

Overview

Relationship

The demo layer over Face Blurring

Blurify.AI's own README points back to the research repo directly: "Check the GitHub repo for building the models, evaluations, etc." Face Blurring is where the YOLOv8n model was trained from scratch and evaluated against Haar Cascade; Blurify.AI is the deployable wrapper — a Streamlit app plus a Dockerfile — with none of the training code.

How it works

Two backends, one blur step

The YOLOv8n path runs ultralytics inference and applies cv2.GaussianBlur((25,25), 100) to each detected box. The Haar Cascade path runs OpenCV's frontal-face classifier (scaleFactor 1.1, minNeighbors 3) and applies a slightly lighter cv2.GaussianBlur((25,25), 75). Both render an original-vs-blurred side-by-side result in the browser.

How it works

A thin Streamlit UI wraps the two detectors trained and evaluated in the Face Blurring research repo — pick a model, upload an image, get a blurred result back.

Upload
image via Streamlit uploader
Pick model
YOLOv8n or Haar Cascade
Blur
GaussianBlur per face box

Install & use

Local
$ git clone https://github.com/baselhusam/Blurify.AI.git
$ pip install -r requirements.txt
$ streamlit run main.py
Docker
$ docker build -t blurify .
$ docker run -p 8501:8501 blurify

Built with

Streamlit
upload + model picker UI
Ultralytics YOLOv8n
one of two blur backends
OpenCV Haar Cascade
the other blur backend
Docker
containerized deployment