Real-time face detection demo running on a Microchip SAMA7G54 evaluation kit. The system captures live video input, identifies human faces, and reports their locations within each video frame.
- Features
- Hardware Requirements
- Software Requirements
- Installation
- Usage
- Troubleshooting
- Version History
- Authors
- License
- Captures video input from a USB camera.
- Performs real-time face detection on the SAMA7G54 MPU.
- Outputs face bounding boxes or similar metadata.
- Optimized for embedded performance.
- Microchip SAMA7G54 Evaluation Kit :contentReference[oaicite:0]{index=0}
- USB camera compatible with the board :contentReference[oaicite:1]{index=1}
- Host PC or terminal emulator to view debugging/log output via serial or USB connection :contentReference[oaicite:2]{index=2}
- Boot firmware / image built for SAMA7G54 that includes support for USB camera input and face detection inference (e.g. Edge Impulse image / model) :contentReference[oaicite:3]{index=3}
- Terminal emulator (e.g. TeraTerm) for console access :contentReference[oaicite:4]{index=4}
- Appropriate drivers / power supply for the board as per Microchip’s user guide :contentReference[oaicite:5]{index=5}
- Follow this guide provided by Edge Impulse to flash the image on your Microprocessor.
- Unpack the SAMA7G54 evaluation board, ensuring you observe ESD precautions.
- Check and set default jumper settings.
- Connect USB-micro-AB cable to the J24 (J-Link-OB USB) port.
- Connect the other end of that USB cable to your PC.
- Open a terminal emulator with settings
115200, N, 8, 1. - Power the board either by:
a. USB-micro-AB cable into J7 (USB-A port), or
b. AC/DC wall adapter into the DC jack connector. - Reset the board. You should see a startup log on the console.
- Flash the face detection demo image (e.g. using Edge Impulse provided image) onto the SAMA7G54 according to the “Primary User Guide” (see page 66). :contentReference[oaicite:6]{index=6}
- After flashing and booting, ensure the USB camera is connected and recognized.
- The demo will start running automatically (or via a command) and begin processing frames.
- Monitor the console output (serial) to see detection results (e.g. coordinates of bounding boxes or number of faces).
- Use any additional provided UI or indicators (if applicable) to view the face detection overlay.
| Problem | Possible Cause | Solution |
|---|---|---|
| No console output | Board not powered / wrong serial settings / wrong port | Check connections, serial settings, and power source |
| USB camera not detected | Incompatible camera / insufficient power / driver support missing | Try a different camera, ensure USB power is adequate, verify driver support |
| Faces not detected or inaccurate | Model not well trained / lighting / camera angle | Improve lighting, adjust camera angle, retrain or fine-tune model |
- v0.1 – Initial release. Basic face detection working.
- v0.2 – Bug fixes and performance optimizations.
- Microchip Technology, Inc.
- For questions please reach out to [email protected]
- https://spdx.org/licenses/BSD-3-Clause-Clear.html
- COCO dataset © 2014 Microsoft, licensed under CC BY 4.0.
- MobileNet model © Google, licensed under Apache 2.0.
- Edge Impulse SDK and generated library, licensed under Apache 2.0.
- Microchip SAMA7G54 evaluation hardware subject to Microchip Technology terms.
- Edge Impulse for image/model generation tools
- Microchip SAMA7G54 documentation for hardware setup
- Open source face detection libraries / prior art