Apppublic

Sudhanshu2601/u-net-crack-detection

sourceHugging Faceupdated 5mo agoView on Hugging Face

0likes

App README

Crack Detection UI

A Flask + React application for detecting cracks in uploaded video footage with a TensorFlow segmentation model. The backend processes each frame, overlays detected cracks with severity colors, and returns a browser-playable WebM output. The React UI shows upload state, model progress, input preview, detected output preview, and download controls.

Live Demo

Open the deployed app on Hugging Face Spaces

Project Structure

—main.py - Flask API, model loading, video processing, progress endpoints
—templates/index.html - Flask-rendered upload page
—frontend/ - React/Vite frontend
—requirements.txt - Python dependencies

Model File

The trained model file is intentionally not tracked in Git because it is large. Place it in the project root with this name before running the backend:

text

crack_segmentation_final.h5

For deployment, the app can download the model from Hugging Face with MODEL_URL:

text

https://huggingface.co/Sudhanshu2601/Cracksegementationdataset/resolve/main/crack_segmentation_final%20(2).h5

Backend Setup

bash

pip install -r requirements.txt
python main.py

The Flask API runs at:

text

http://127.0.0.1:5000

Frontend Setup

bash

cd frontend
npm install
npm run dev

The React UI usually runs at:

text

http://127.0.0.1:5173

Notes

Generated uploads, processed videos, CSV reports, frontend build files, dependencies, and model weights are excluded from Git.

Free Deployment on Hugging Face Spaces

1.Create a new Space at https://huggingface.co/new-space.
2.Choose Docker as the Space SDK.
3.Choose Public visibility.
4.Use the free CPU Basic hardware.
5.Push this repository's files to the Space repository.
6.The app will use MODEL_URL to download the model during startup.

The first build/start can take several minutes because the container installs TensorFlow and downloads the model file.