# handle_rec

**Repository Path**: ausmith/handle_rec

## Basic Information

- **Project Name**: handle_rec
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2026-02-18
- **Last Updated**: 2026-02-18

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Hand Recognition Prototype

Web-based prototype for real-time hand ROI detection, approximate keypoints, and lightweight gesture classification.

## Quick Start

1. Ensure local OpenCV runtime files exist:

```text
vendor/opencv.js
vendor/opencv.wasm
```

2. Start a local server (camera access requires HTTPS or localhost):

```bash
python3 -m http.server 8000
```

3. Open:

```text
http://localhost:8000
```

4. Allow camera permission when prompted.

## OpenCV.js Local Setup (Required)

This app **only** loads `./vendor/opencv.js`. CDN is not used.

Steps:
1. Download `opencv.js` from the official OpenCV.js build.
2. Download matching `opencv.wasm` from the same build.
3. Save both in the project root `vendor/` directory.
4. Run the local server and reload.

Optional: override via `window.OPENCV_JS_URL` if needed.

## Troubleshooting

- `OpenCV.js init timeout (check vendor/opencv.wasm)`:
  - Verify `vendor/opencv.wasm` exists.
  - Verify Network panel shows `200` for `/vendor/opencv.wasm`.
- `Camera permission denied...`:
  - Open site permissions and allow camera, then reload.
- `No camera found...`:
  - Connect/enable camera device, then reload.
- `Camera access requires HTTPS or localhost.`:
  - Use `http://localhost:8000` or an HTTPS origin.

## Notes
- Keypoints are approximated from contour and convex hull; replace with a trained model later.
- Designed to keep the pipeline and data schema compatible with embedded constraints.