Bureaucrats, dave_user, Administrators
8,156
edits
Changes
no edit summary
{{InfoBoxTop}}
{{AppliesToAxel}}
{{AppliesToAxelEsatta}}
{{AppliesToAxelLite}}
{{AppliesToAxelEsattaAppliesToAXEL Lite TN}}{{AppliesToSBCX}}{{AppliesToIoT}}{{AppliesToMachineLearning}}
{{InfoBoxBottom}}
{{WarningMessage|text=This technical note was validated against specific versions of hardware and software. It may not work with other versions.}}
|-
|1.0.0
|Ocotber October 2019
|First public release
|}
==Introduction==
Nowadays, Machine Learning (ML) and Deep Learning (DL) technologies are getting popular in the embedded world as well. Several different approaches are available to deploy such technologies on embedded devices. This Technical Note (TN) describes such an approach, which makes use of a Tensor Flow model generated with [https://www.customvision.ai Microsoft Azure Custom Vision service]. ==Testbed basic configuration==The testbed consists of an [[:Category:SBC-AXEL|SBCX Single Board Computer]] equipped with an i.MX6Q-powered [[:Category:AxelLite|Axel Lite]] system-on-module (SoM).
==Testbed configurationTest application==The testbed consists of an test application is a classical image classifier. The following classes are supported:<pre>AvocadoBananaGreen AppleHandOrangeRed Apple</pre> The following image shows the application's architecture. [[File:Category:SBCSBCX-image-AXELclassifier.png|SBCX Single Board Computerthumb|center|600px|Test application's architecture]] equipped It mainly consists of the following blocks:* The top-level application code (Python)* The TensorFlow platform* The TensorFlow model* The OpenCV library. As stated in the introduction, the classifier is based on a model that was generated with Azure Custom Vision. In particular, the model was retrieved from [https://dev.to/azure/creating-an i.MX6Q-powered [image-recognition-solution-with-azure-iot-edge-and-azure-cognitive-services-4n5i this project] by [https:Category:AxelLite|Axel Lite//dev.to/gloveboxes Dave Glover]] system. Glover's project is extremely useful to understand how Custom Vision—and, in general, Azure Cognitive Services—work. It is worth remembering that '''no particular Machine Learning-related skills are required to create such a model'''. Glover's project follows the approach suggested by Azure, which makes use of containers. For the sake of simplicity, this Technical Note is based on-module (SoM)a simpler strategy, which is closer to the usual approach used in the embedded world. As such, it doesn't make use of any container.
TensorFlow release is 2.0.0. TensorFlow It was built with the proper configuration for the i.MX6 SoC. For more details, please refer to [https://github.com/lhelontra/tensorflow-on-arm this project] by [https://github.com/lhelontra Leonardo Lontra].
Last but not least, the application code is based on [https://docs.microsoft.com/en-us/azure/cognitive-services/custom-vision-service/export-model-python this example].
==Performances==
The following box shows the output of the application while classifying an image that contains a red apple:
<pre class="board-terminal">
$ python3 image-classifier.py
2019-10-25 11:17:15,288 - DEBUG - Starting ...
2019-10-25 11:17:15,289 - DEBUG - Importing the TF graph ...
Classified as: Red Apple
2019-10-25 11:17:21,591 - DEBUG - Prediction time = 2.567471504211426 s
Avocado 2.246000076411292e-05
Banana 3.769999921132694e-06
Green Apple 0.029635440558195114
Hand 4.4839998736279085e-05
Orange 0.0009084499906748533
Red Apple 0.9693851470947266
2019-10-25 11:17:21,594 - DEBUG - Exiting ...
</pre>
The image was classified as a "Red Apple" because the probability associated with this class was by far the highest (almost 97%).
[[File:Red-apple.jpg|thumb|center|300px|The image shown in the previous example]]
During the execution of the test application, the status of the processes and the ARM cores was observed with the help of the <code>htop</code> tool.
[[File:SBCX-image-classifier-1.png|thumb|center|600px|<code>htop</code> during the execution of the test application]]
By default, the scaling governor is set to interactive:
<pre class="board-terminal">
root@sbcx:~# cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
interactive
</pre>
Therefore, during the execution of the program, the cores' frequency was scaled to 1 GHz as expected.
At first glance, it seemed that the TensorFlow platform does not exploit the available cores to implement significant parallel computing. To verify this guessing, the test was repeated by limiting the number of cores. The results are listed in the following table.
{| class="wikitable"
|+
!Platform
!# of cores
!Governor
!Prediction time
[s]
|-
|SBCX (i.MX6Q)
|1
|interactive
|3.0
|-
|SBCX (i.MX6Q)
|2
|interactive
|2.6
|-
|SBCX (i.MX6Q)
|4
|interactive
|2.5
|-
|SBCX (i.MX6Q)
|4
|userspace
(1 GHz)
|2.5
|}
As shown, the prediction time doesn't change significantly with increasing number of cores. This should confirm TensorFlow doesn't implement a parallel computing engine on this architecture.
