Cat Detector

A simple Progressive Web App that detects cats in images using a lightweight ONNX model running entirely in the browser.

Usage

Open frontend/index.html locally or visit the GitHub Pages site. The application installs as a PWA and works offline after the first visit. Upload an image and click Predict to see if a cat is detected.

Development

No backend or Docker is required. The app loads squeezenet1_1.onnx with ONNX Runtime Web.

Deployment

GitHub Actions automatically publishes the frontend directory to the gh-pages branch.

Large files

Sample test images are tracked with Git LFS. The model weights are stored directly in the repository, so additional commands are not required to run the web application. The test image dataset is omitted from the repository to keep the pull request lightweight.

Post-merge manual steps

1. Заменить *.png.empty настоящими иконками (180 / 192 / 512 px).
2. Запустить bash scripts/gen-splash.sh и закоммитить PNG из папки splash/.
3. Проверить PWA на iPhone: отсутствие авто-зумов, корректная safe-area, launch-screen.