Edge Impulse Docs

Edge Impulse Documentation

Welcome to the Edge Impulse documentation. You'll find comprehensive guides and documentation to help you start working with Edge Impulse as quickly as possible, as well as support if you get stuck. Let's jump right in!

On your Arduino

Impulses can be deployed as an Arduino library. This packages all your signal processing blocks, configuration and learning blocks up into a single package. You can include this package in your own sketches to run the impulse locally. In this tutorial you'll export an impulse, and integrate the impulse in a sketch to classify sensor data.

This tutorial should work on most Arm-based Arduino development boards with at least 64K of RAM, like the Arduino Nano 33 BLE Sense.

📘

Knowledge required

This tutorial assumes that you're familiar with the Arduino IDE, and that you're comfortable building Arduino sketches. If you're unfamiliar with these tools you can build binaries directly for your development board from the Deployment page in the studio.

Prerequisites

Make sure you followed the Continuous motion recognition tutorial, and have a trained impulse. Also install the following software:

And make sure you can compile sketches for your development board.

Deploying your impulse

Head over to your Edge Impulse project, and go to Deployment. From here you can create the full library which contains the impulse and all external required libraries. Select Arduino library and click Build to create the library. Then download and extract the .zip file.

Then to add the library and open an example, open the Arduino IDE and:

  1. Choose Sketch > Include Library > Add .ZIP library....
  1. Find the folder (do not go inside the folder), and select Choose.
  1. Then, load an example by going to File > Examples > Your project name - Edge Impulse > static_buffer.
  1. Voila. You now have an example application that loads your impulse.

Running the impulse

With the project ready it's time to verify that the application works. Head back to the studio and click on Live classification. Then load a validation sample, and click on a row under 'Detailed result'.

Selecting the row with timestamp '320' under 'Detailed result'.

To verify that the local application classifies the same, we need the raw features for this timestamp. To do so click on the 'Copy to clipboard' button next to 'Raw features'. This will copy the raw values from this validation file, before any signal processing or inferencing happened.

Copying the raw features

In the sketch paste the raw features inside the static const float features[] definition, for example:

static const float features[] = {
    -19.8800, -0.6900, 8.2300, -17.6600, -1.1300, 5.9700, ...
};

Then click Upload in the Arduino IDE to build and flash the application.

Seeing the output

To see the output of the impulse, open the serial monitor from the Arduino IDE via Tools > Serial monitor, and selecting baud rate 115,200.

This will run the signal processing pipeline, and then classify the output:

Edge Impulse standalone inferencing (Arduino)
Running neural network...
Predictions (time: 0 ms.):
idle:   0.015319
snake:  0.000444
updown: 0.006182
wave:   0.978056
Anomaly score (time: 0 ms.): 0.133557
run_classifier_returned: 0
[0.01532, 0.00044, 0.00618, 0.97806, 0.134]

Which matches the values we just saw in the studio. You now have your impulse running on your Arduino development board!

Troubleshooting

Multiple libraries were found for ...

Exported libraries are automatically versioned, but it's possible that the Arduino IDE gets confused on which version to use leading to an error like: Multiple libraries were found for .... You can delete old versions of the libraries to mitigate this. The libraries are located at:

  • Windows: My Documents > Arduino > libraries
  • macOS: ~/Documents/Arduino/libraries/
  • Linux: ~/sketchbook/libraries

error: reference to 'SerialUSB' is ambiguous

This error shows up on the Arduino Nano 33 BLE Sense when using the Arduino Core v1.1.6 for the target. We've filed a bug with Arduino (here) but until that is resolved you can revert the Arduino core version back to v1.1.4 to compile your sketch.

Selecting the 1.1.4 Arduino core for the Arduino Nano 33 BLE Sense

No such file or directory: include <arm_math.h>

~/Documents/Arduino/libraries/ei-accelerometer-impulse/src/edge-impulse-sdk/CMSIS/NN/Source/PoolingFunctions/arm_max_pool_s8.c:32:10: fatal error: arm_math.h: No such file or directory
 #include <arm_math.h>

Some users have indicated that this issue can be solved by reinstalling the Arduino IDE.

macro "min" passed 3 arguments, but takes just 2

If you're compiling on a SAMD21-based target, you'll see the above error. This is a known bug in the Arduino core for the SAMD21. If you apply the following patch the issue will go away.

Empty array when printing results

If the predictions are not properly printed, e.g.:

Edge Impulse standalone inferencing (Arduino)
run_classifier returned: 0
Predictions (DSP: 369 ms., Classification: 3 ms., Anomaly: 4 ms.): 
[, , , , ]

Then your printing library does not support printing floating point numbers (this happens for example on the Arduino MKR WAN 1300). You can get around this by converting the predictions to integers. E.g.:

ei_printf("Predictions (DSP: %d ms., Classification: %d ms., Anomaly: %d ms.): \n",
        result.timing.dsp, result.timing.classification, result.timing.anomaly);

    // print the predictions
    ei_printf("[");
    for (size_t ix = 0; ix < EI_CLASSIFIER_LABEL_COUNT; ix++) {
        ei_printf("%d", static_cast<int>(result.classification[ix].value * 100));
#if EI_CLASSIFIER_HAS_ANOMALY == 1
        ei_printf(", ");
#else
        if (ix != EI_CLASSIFIER_LABEL_COUNT - 1) {
            ei_printf(", ");
        }
#endif
    }
#if EI_CLASSIFIER_HAS_ANOMALY == 1
    ei_printf("%d", static_cast<int>(result.anomaly));
#endif
    ei_printf("]\n");

Slow DSP operations

Where possible the signal processing code utilizes the vector extensions on your platform, but these are not enabled on all platforms. If these are not enabled we fall back to a software implementation which is slower. We don't enable these on all platforms because the wide variety of platform and core versions Arduino supports, but you can see enable them for your platform by adding the following code on the first line of your sketch, before any includes (only works on Arm cores):

#define EIDSP_USE_CMSIS_DSP             1
#define EIDSP_LOAD_CMSIS_DSP_SOURCES    1

If this still does not work you need to edit the src/edge-impulse-sdk/dsp/config.hpp file in the library and add, before #ifndef EIDSP_USE_CMSIS_DSP:

#define EIDSP_USE_CMSIS_DSP             1
#define EIDSP_LOAD_CMSIS_DSP_SOURCES    1

If you have a target that ships on old versions of CMSIS Core (like the SAMD21 targets) you might need to also declare some new macros, e.g.:

#define __STATIC_FORCEINLINE                   __attribute__((always_inline)) static inline
#define __SSAT(ARG1, ARG2) \
__extension__ \
({                          \
  int32_t __RES, __ARG1 = (ARG1); \
  __ASM volatile ("ssat %0, %1, %2" : "=r" (__RES) :  "I" (ARG2), "r" (__ARG1) : "cc" ); \
  __RES; \
 })

Code compiling fails under Windows OS

fork/exec C:\Users\MYUSER\AppData\Local\Arduino15\packages\arduino\tools\arm-none-eabi-gcc\7-2017q4/bin/arm-none-eabi-g++.exe: The filename or extension is too long.

This error is usually thrown when the list of object files to compile exceeds Windows max number of characters (32k) in a command line. To overcome this issue, download the platform.local.txt file and copy it under your mbed directory, ie:

C:\Users\MYUSER\AppData\Local\Arduino15\packages\arduino\hardware\mbed\1.1.4\

Updated 19 days ago

On your Arduino


Suggested Edits are limited on API Reference Pages

You can only suggest edits to Markdown body content, but not to the API spec.