28 réponses à ce sujet

[electroniclinic]

[electroniclinic]

I turned the RDK X5 into a personal gym trainer - it watches my body,
tracks my movement, and counts my reps automatically. No smartwatch,
no phone, just pure computer vision.

But here's the fun part I didn't just build ONE version. I tested
**3 completely different AI methods** to see which one actually holds up:

**Method 1 - Bounding Box (YOLO):** simple, beginner-friendly, tracks
head movement. Easy to build, but breaks easily if you move sideways.

**Method 2 - Body Landmarks (MediaPipe):** tracks real knee angles
for way better accuracy, but it's heavy on the CPU and can lag on
embedded hardware.

**Method 3 - BPU-Powered AI Model:** runs directly on the RDK X5's
hardware AI accelerator. Fast, stable, and accurate enough for an
actual smart-gym product.

The results genuinely surprised me - go watch and see which one you
would've picked before the reveal

[drobo_community]

Great work!Please post in the #rdk-project-showcase channel.

[electroniclinic]

Hey everyone!
New video just dropped - this one's about building an AI CCTV system that decides on its own when to record
I used the RDK X5 with YOLOv5 to build two versions:
A manual CCTV system with real-time AI person detection
A fully automatic system that only starts recording when someone actually steps into a defined zone and stays there - no false triggers, no random motion alerts
Goes into region-of-interest detection, glitch protection, time-based confirmation logic, and GPIO LED feedback - basically how real smart surveillance systems make decisions.
Would love to know what you all think, and if you build your own version, drop it here!

[electroniclinic]

[electroniclinic]

Hi everyone!

My latest project is finally live! In this video, I show how to combine the RD-03D mmWave Radar, an ESP32-C3, and the D-Robotics RDK X5 to build several practical and interactive applications.

In this video, I demonstrate:
Real-time radar tracking
Secure-zone intrusion detection
Radar-triggered USB camera
Python applications running on the RDK X5
Android radar app
A Snake game controlled entirely by body movement
Android version of the radar-controlled Snake game

This project shows how easy it is to use an ESP32-C3 to handle the radar communication, allowing the RDK X5 to focus on running higher-level applications.

If you're interested in RDK X5, Edge AI, Python, computer vision, or mmWave radar, I'd love to hear your thoughts and feedback.

Thank you to the D-Robotics team and this community for building such a great platform for developers. I hope you enjoy the project!

[electroniclinic]

[electroniclinic]

New RDK X5 project: an AI assistant that wakes up on its own

Just finished a build that pairs mmWave radar with ChatGPT (GPT-4o), running entirely on the RDK X5. No wake word, no button - the moment the radar detects someone in the room, the RDK X5 triggers the assistant and it starts talking.

Once active, it uses the RDK X5's USB camera input to capture and send frames to GPT-4o for real scene understanding - not limited to a fixed object list like COCO. It correctly picked up on someone hiding behind a cupboard just out of direct view, which honestly surprised me too.

The RDK X5 is also handling GPIO output in this build - voice commands control an LED directly through the board, so it's not just a chatbot, it's actually driving hardware in real time.

Stack:
RDK X5 - core processing, camera input, GPIO control
RD-03D mmWave radar + Xiao ESP32-C3 - presence trigger
GPT-4o - conversation + vision
ElevenLabs - voice output

[electroniclinic]