Start here
Understand the preview and choose your path.
HealthCam connects a local computer vision pipeline to a cloud application. The agent captures webcam segments, runs enabled analyzers on your machine, and uploads structured results for storage and dashboards.
These guides are for developers and technical testers. They describe the current repository, including the manual setup and unfinished integrations.
The documentation and hosted demo are public. The source repository is currently private, so installing the agent requires GitHub repository access.
Choose your path
| Your goal | Start with | What you need |
|---|---|---|
| See the interface | Explore the hosted demo | A browser; choose Explore the demo |
| Test capture and inference | Install the local agent | Git, Python, Docker, FFmpeg, a webcam |
| Generate heart-rate dashboard data | Install the agent, with the BPM analyzer enabled | A compatible NVIDIA GPU and container runtime |
| Work on the web application | Run the web app locally | Node.js, npm, PostgreSQL |
| Add a computer vision capability | Build an analyzer | A working local agent and an analyzer implementation |
The shared demo is read-only and uses synthetic telemetry. Use a personal Google account for your own ingestion key and uploaded results.
What works today
- Manifest-driven installation and booting of configured local services.
- Webcam capture and active-application context collection.
- Parallel execution of enabled analyzers on the same video segment.
- A pyVHR heart-rate analyzer and a MediaPipe-based blink analyzer.
- Authenticated cloud ingestion, raw telemetry storage, and scheduled BPM aggregation.
- A BPM dashboard, catalogue pages, and local service status display.
What is still unfinished
There is no packaged, ready-to-install product release. You currently clone the GitHub repository, install prerequisites, edit configuration, and run the boot script.
The Store records plugin and dashboard ownership. It does not download and install services on your computer. Local services are enabled in config.services.json. The blink analyzer produces results, but a complete blink dashboard and alert experience are still in progress. Older tray launchers target an earlier architecture and are not the installation path in these guides.
Before collecting data
Video stays on your computer. The dispatcher deletes each segment and its metadata file after successful analysis and upload. Failed processing or uploads may retain local files.
Uploaded telemetry includes measurements, application/window context, timestamps, and file metadata. Start with non-sensitive desktop content and a short test. Local APIs currently lack authentication and should remain on a trusted development machine, without port forwarding or public exposure.
HealthCam is an experimental thesis and portfolio project, not a medical device. The original rPPG validation did not establish reliable discrimination between application-level heart-rate differences. See the validation evidence.
Follow the installation-to-test sequence
- Check the requirements for your operating system and chosen analyzer.
- Clone the repository and prepare a Python environment.
- Set your cloud endpoint and personal ingestion key.
- Enable services and boot the agent.
- Run a short recording and verify each stage in your first test.
A successful test means capture, inference, upload, and cleanup are observable. It does not by itself validate the accuracy of an analyzer.