DFRobot’s ESP32-S3 Edge AI Modules Combine Voice, Vision & Gesture Recognition Without the Cloud

DFRobot’s ESP32-S3 Edge AI Modules

The Gravity sensor detects 5 hand gestures and up to 10 faces or upper-body presences within 3 meters, while the ESP32-S3 AI CAM adds voice control, night vision, and Wi-Fi/BLE connectivity, making them useful for privacy-focused IoT use such as smart homes, kiosks, or automation systems.

The Gravity sensor runs AI locally with I2C/UART support, 3.3–5V input, 100mA typical draw, and real-time RGB gesture indicators. The ESP32-S3 CAM features a 3MP IR wide-angle camera, onboard mic, speaker amp, ChatGPT voice interaction, and supports EdgeImpulse, YOLO, and OpenCV. It includes IR LEDs, a light sensor, an SD slot, and runs on 3.3V or 5–12V via USB/Type-C. Both modules offer fast, local AI processing without cloud reliance.

Offline Edge AI Gesture & Face Detection Sensor

ESP32-S3 AI Camera Module Specifications

ESP32-S3 Edge AI Modules Specifications

• Gravity: Offline Edge AI Gesture & Face Detection Sensor
  • MCU – Espressif ESP32-S3R8 dual-core Xtensa LX7 processor with AI acceleration
  • Recognition Capabilities
    • Gesture Recognition: 5 predefined hand gestures
      • Thumbs up
      • Index + middle fingers
      • ️ Open palm
      • Thumb + little finger
      • Middle, ring, and little fingers
    • Face and Upper Body Detection
    • Recognition Range:
      • Gesture: 0.5 to 3 meters
      • Face/Body: 0.5 to 3 meters
    • Field of View (FoV): 85° diagonal
    • Camera Focal Length: 1.56mm
    • Position Coordinate Output Supported
  • Connectivity
    • Communication Interfaces: I2C and UART
    • UART Protocol: Modbus (RS-485)
    • I2C Address: 0x72
    • UART Default Baud Rate: 9600bps
    • Connector Type: PH2.0-4P and 2.54mm header
    • Interrupt Output: 2.54mm pin header
  • Misc
    • RGB Gesture Indicator:
      • Blue: Thumbs up
      • Green: Middle, ring, little fingers
      • Red: Open palm
      • Yellow: Index + middle fingers
      • Purple: Thumb + little finger
      • Presence Detection LED
  • PCB Dimensions – 42mm × 32mm
  • Mounting Hole Spacing: 25mm × 35mm
  • Mounting Hole Diameter: 3.1mm
• ESP32-S3 AI Camera Module Specifications
  • SoC: Espressif ESP32-S3R8 dual-core Xtensa LX7 processor with AI acceleration
  • Storage:
    • Flash: 16MB NOR
    • MicroSD card (Max size not specified in datasheet)
  • Connectivity
    • Wi-Fi 4 (802.11 b/g/n) and Bluetooth 5.0 LE
  • Camera Sensor: OV3660 3MP wide-angle IR camera
    • Field of View: 160°
    • Focal Length: 0.95mm
    • Aperture: f/2.0 ±5%
    • IR Sensitivity: 940nm
    • Distortion: <8%
  • AI and Voice Features
    • Edge AI support (Edge Impulse, YOLOv5, OpenCV)
    • Online AI model access with ChatGPT integration
    • Built-in I2S PDM microphone and MAX98357 I2S amplifier
    • Voice interaction and ASR (automatic speech recognition) support
  • Night Vision
    • Infrared (IR) LED illumination (IO47)
    • LTR-308 ambient light sensor for brightness detection
  • USB – 1x USB Type-C for power and programming
  • Serial – Gravity connector (UART/I2C)
  • Misc
    • MX1.25-2P speaker interface
    • Onboard LED (IO3)
    • Reset and BOOT buttons
  • Power
    • Operating Voltage – 3.3V
    • VIN Input – 5-12V DC
    • Type-C Input – 5V DC
    • Power Chip – HM6245
  • Dimensions: 42 x 42 mm
  • Temperature: -10°C to +60°C

Both the sensor and the camera module offer software support for AIoT projects. The gesture sensor enables development via MakeCode, Mind+, and Arduino libraries, with protocol docs for accessing gesture IDs, face counts, coordinates, and confidence scores. The ESP32-S3 AI Camera supports ESP-IDF, Edge Impulse for on-device AI training, OpenCV for image processing, and ChatGPT-based voice interaction. It includes sample code for camera setup, audio streaming, and cloud AI communication via Wi-Fi, with extensive tutorials and GitHub resources to accelerate development. More information can be found on their respective wiki page linked in the products page.

Previously, we have written about other ESP32-S3 Edge AI modules, including the M5Stack StampS3A, the Waveshare ESP32-S3 1.8-inch Knob display, and Adafruit Sparkle Motion. These devices are built around the ESP32-S3 and can run an edge AI application very easily.

The ESP32-S3 AI Camera Module and the Gravity: Offline Edge AI Gesture & Face Detection Sensor are available on dfrobot’s store page, where they cost around $14.90 and $18.90, respectively. Additional information, including code examples and videos, can be found on the same page.

Images used courtesy of DFrobot