Failed to get information off device Sensor readings from device do not seem to be all numbers

Hi, I’ve been trying to connect my AI-Thinker Esp32 to my project, and tried following the instructions given, but I’m getting this error: “Failed to get information off device Sensor readings from device do not seem to be all numbers, found: [“Timing:”,“DSP”,“8”,“ms”,“inference”,“1210”,“ms”,“anomaly”,“0”,“ms”]”.

It’s my first time working with a microcontroller, so yeah, kinda don’t what I really need to do sometimes. If anyone could help me, I’d appreciate. Thank you.

Since you are new to this stuff I would follow the EdgeImpulse ESP tutorial.

  • What sensor is you model making predictions on (camera, microphone, accelerometer, etc.)?
  • What is the link to the tutorial you are using?
  • How are you trying to connect the ESP to your project (WebUSB, CLI, etc.)?

Hi, I’ve tried following that tutorial you sent, as well as this one: Arduino library - Edge Impulse Documentation

I’m trying to use the camera, and connecting via CLI.
I tried connecting again and got a slightly different error now: “Failed to get information off device Sensor readings from device do not seem to be all numbers, found: [“garrafa:”,“0.01562”]”
Garrafa means bottle.

I also tried following this video: Simple ESP32-CAM Object Detection - YouTube
But got the same results.

Before running the CLI to connect the ESP32 to the Studio are you flashing the firmware on to the ESP32 using flash_windows.bat? If so, then the ESP32 should connect to the Studio. However, the AIThinker ESP32 is not an officially supported board so it may not work. The ESP world is so frustrating and confusing.

If you can’t get the ESP32 connected via the EdgeImpulse CLI, then save images from the ESP32 to an SD card and upload the JPGs or PNGs via the Studio Data Acq page.

Also, you do not need to uses images captured by the ESP32 to train a model. The training images can come from anywhere like captured with your phone or digital camera.

I tried using the flash_windows.bat but eventually gave up because for what I understood, it doesn’t work with the AIThinker ESP32. So, been trying with Arduino IDE. I used the phone camera for training the model, but now I wanted to try it with the ESP32, and see how the live classification would work with it.

You best option is to deploy from the Studio either Arduino Library or the C++ Library and integrate the Library with your code that reads the camera.

1 Like

I haven’t really had any progress on my problem and I just noticed your answer now, but I don’t know if I understand what you’re saying. I tried using the Arduino Library option when deploying in my Edge Impulse project, if that’s what you’re suggesting.

  • When you deploy the your Impulse as an Arduino Library and install the ZIP file via the Arduino IDE there will be example ESP examples INO files installed.
  • Go under File > Examples > Examples for Custom Libraries > YourLibraryName > esp > esp_camera.
    • This will open a esp_camera.ino file.
    • If this will not run on your ESP, then you’ll need to modify it for you circuit board and pinout.
      • For example the following will need to be modified in esp_camera.ino:
        • the GPIO pins at the top of the code
        • Modify bool ei_camera_init() and you’ll probably have to write your own esp_camera_sensor_get().
        • Modify void ei_camera_deinit()
        • Modify bool ei_camera_capture() and the framebuffer reader esp_camera_fb_get();.
        • Once you get your buffers filled, static int ei_camera_get_data() may work as-is.
  • To make the code modifications based on your MCU, I would find a simple program that can read the camera on your board.
    • Then merge the 2 programs as appropriate.

Yeah, that’s a lot of work and now you can truly appreciate the work EdgeImpulse puts into making the various MCU an officially supported board. And most of the EI code is open source, so try diving into their GitHub repository so you can fork for your hardware.