Last edited 9 months ago

X-LINUX-AI application samples zoo: Difference between revisions

Visual
Registered User
mNo edit summary
Registered User
mNo edit summary
Tag: 2017 source edit
 
(27 intermediate revisions by 6 users not shown)
Line 1: Line 1:
<noinclude>{{ApplicableFor
{{ApplicableFor
|MPUs list=STM32MP13x, STM32MP15x
|MPUs list=STM32MP13x, STM32MP15x, STM32MP25x
|MPUs checklist=STM32MP13x,STM32MP15x
|MPUs checklist=STM32MP13x, STM32MP15x, STM32MP25x
}}</noinclude>
}}
All X-LINUX-AI application samples are listed in this page.
==Application samples==
 
<noinclude></noinclude>
==TensorFlow Lite application samples==
{| class="st-table" style="text-align:center; width:100%; border-style: hidden; margin-left: 0px"
{| class="st-table" style="text-align:left; margin-left: 0px"
! style="border-width: 1px; border-style: solid; border-color: #f3f6f4" |
| style="border-style: hidden" |
! style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | [[File:X-LINUX-AI_Machine_code.png|center|150px|link=]] Only for STM32MP2<sup> 1</sup>
[[File:X-LINUX-AI_TFLite_cpp.png|75px|link=]]
! style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | [[File:X-LINUX-AI_TFLite.png|center|150px|link=]]
| style="border-style: hidden" |
! style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | [[File:X-LINUX-AI_Onnx.png|center|150px|link=]]
* [[X-LINUX-AI_-_face_recognition_using_TensorFlow_Lite_C%2B%2B_API | Face recognition using TensorFlow Lite C++ API]]
! style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | [[File:X-LINUX-AI_coral.png|center|75px|link=]]
* [[X-LINUX-AI_-_image_classification_using_TensorFlow_Lite_C%2B%2B_API | Image classification using TensorFlow Lite C++ API]]
|- style="height:50px; vertical-align:middle;border-width: 1px; border-style: solid; border-color: #f3f6f4"
* [[X-LINUX-AI_-_object_detection_using_TensorFlow_Lite_C%2B%2B_API | Object detection using TensorFlow Lite C++ API]]
! style="border-width: 1px; border-left-style: hidden; border-top-style: hidden;border-color: #f3f6f4"|
|-
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{Highlight|NPU/GPU}}   
| style="border-style: hidden" |
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | '''CPU'''
[[File:X-LINUX-AI_TFLite_python.png|75px|link=]]
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | '''CPU'''
| style="border-style: hidden" |
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | '''EdgeTPU'''
* [[X-LINUX-AI_-_image_classification_using_TensorFlow_Lite_Python_runtime | Image classification using TensorFlow Lite Python runtime]]
|- style="height:50px; vertical-align:middle;border-width: 1px; border-style: solid; border-color: #f3f6f4"
* [[X-LINUX-AI_-_object_detection_using_TensorFlow_Lite_Python_runtime | Object detection using TensorFlow Lite Python runtime]]
! style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | [[File: Image_classification_logo.png|150px|link=Image_classification]]<br />Image classification
|}
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{Y}}     
 
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{Y}}
==Coral Edge TPU application samples==
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{Y}}
{| class="st-table" style="text-align:left; margin-left: 0px"
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{Y}}
| style="border-style: hidden" |
|- style="height:50px; vertical-align:middle;border-width: 1px; border-style: solid; border-color: #f3f6f4"
[[File:X-LINUX-AI_TFLite_EdgeTPU_cpp.png|75px|link=]]
! style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | [[File: Object_detection_logo.png|150px|link=Object_detection]]<br />Object detection
| style="border-style: hidden" |
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{Y}}
* [[X-LINUX-AI_-_image_classification_using_Coral_Edge_TPU_TensorFlow_Lite_C%2B%2B_API | Image classification using Coral Edge TPU TensorFlow Lite C++ API]]
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{Y}}
* [[X-LINUX-AI_-_object_detection_using_Coral_Edge_TPU_TensorFlow_Lite_C%2B%2B_API | Object detection using using Coral Edge TPU TensorFlow Lite C++ API]]
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{Y}}
|-
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{Y}}
| style="border-style: hidden" |
|- style="height:50px; vertical-align:middle;border-width: 1px; border-style: solid; border-color: #f3f6f4"
[[File:X-LINUX-AI_TFLite_EdgeTPU_python.png|75px|link=]]
! style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | [[File: pose_estimation_logo.png|150px|link=Pose_estimation]]<br />Pose estimation
| style="border-style: hidden" |
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{Y}}
* [[X-LINUX-AI_-_image_classification_using_Coral_Edge_TPU_TensorFlow_Lite_Python_runtime | Image classification using Coral Edge TPU TensorFlow Lite Python runtime]]
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{N}}
* [[X-LINUX-AI_-_object_detection_using_Coral_Edge_TPU_TensorFlow_Lite_Python_runtime | Object detection using using Coral Edge TPU TensorFlow Lite Python runtime]]
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{N}}
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{N}}
|- style="height:50px; vertical-align:middle;border-width: 1px; border-style: solid; border-color: #f3f6f4"
! style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | [[File: semantic_segmentation_logo.png|150px|link=Semantic_segmentation]]<br />Semantic segmentation
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{Y}}
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{N}}
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{N}}
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{N}}
|- style="height:50px; vertical-align:middle;border-width: 1px; border-style: solid; border-color: #f3f6f4"
! style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | [[File: Face_recognition_logo.png|150px|link=]]<br />Face recognition
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{N}}
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{Y}}
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{N}}
| style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | {{N}}
|}
|}
<br>
{{Important|<sup>1</sup> NPU/GPU acceleration based on OpenVX backend using network binary graph (NBG) is available only for STM32MP2x boards}}
==Performance Comparison==
<br>


==armNN application samples==
{| class="st-table" style="text-align:center; width:100%; border-style: hidden; margin-left: 0px"
{| class="st-table" style="text-align:left; margin-left: 0px"
!width="15%"  style="border-width: 1px; border-style: solid; border-color: #f3f6f4" |
| style="border-style: hidden" |
!width="20%"  style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | Model
[[File:X-LINUX-AI_armNN_TFLite_cpp.png|75px|link=]]
!width="35%"  style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | STM32MP25<br>{{Highlight|NPU}}
| style="border-style: hidden" |
!width="30%"  style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | STM32MP15x<br>{{Highlight|CPU}}
* [[X-LINUX-AI_-_image_classification_using_armNN_TensorFlow_Lite_parser | Image classification using armNN TensorFlow Lite parser]]
|- style="height:100px; vertical-align:middle;"
* [[X-LINUX-AI_-_object_detection_using_armNN_TensorFlow_Lite_parser | Object detection using armNN TensorFlow Lite parser]]
! width="15%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" rowspan="2" | [[File: Image_classification_logo.png|100px|link=Image_classification]]<br />Image classification
| width="25%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | MobilenetV1_0.5 128x128
| width="30%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" |
'''515 FPS'''
| width="30%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" |
22 FPS
|- style="height:100px; vertical-align:middle;"
| width="25%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | MobilenetV2_1.0 224x224
| width="30%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" |
'''72 FPS'''
| width="30%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" |
2 FPS
|- style="height:100px; vertical-align:middle;"
! width="15%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" rowspan="2" | [[File: Object_detection_logo.png|100px|link=Object_detection]]<br />Object detection
| width="25%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | SSD_MobilenetV1_1.0 300x300
| width="30%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" |
'''40 FPS'''
| width="30%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" |
2 FPS <br>
|- style="height:100px; vertical-align:middle;"
| width="25%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | SSD_MobilenetV2_FPNLite 224x224
| width="30%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | '''36 FPS'''
| width="30%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | 0.8 FPS
|- style="height:100px; vertical-align:middle;"
! width="15%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | [[File: Face_recognition_logo.png|100px|link=]]<br />Face recognition
| width="25%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | STModel_FaceReco
| width="29%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" |
'''15 FPS'''
| width="29%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" |
5 FPS
|- style="height:100px; vertical-align:middle;"
! width="15%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | [[File: pose_estimation_logo.png|100px|link=Pose_estimation]]<br />Pose estimation
| width="25%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | YoloV8n 256x256
| width="30%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" |
'''59 FPS'''
| width="30%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" |
1.3 FPS <br>
|- style="height:100px; vertical-align:middle;"
! width="15%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | [[File: semantic_segmentation_logo.png|100px|link=Semantic_segmentation]]<br />Semantic segmentation
| width="25%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" | DeepLabV3 257x257
| width="30%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" |
'''17 FPS'''
| width="30%" style="border-width: 1px; border-style: solid; border-color: #f3f6f4" |  
1 FPS <br>
|}
|}
<br>


<noinclude>
<noinclude>
[[Category:Artificial intelligence expansion packages|01]]
[[Category:Sub-articles]]
{{NoIndex}}
{{PublicationRequestId |25195| 2022-11-07 | }}
</noinclude>
</noinclude>

Latest revision as of 18:49, 28 June 2024↑

Applicable for STM32MP13x lines, STM32MP15x lines, Unknown MPU

1. Application samples[edit | edit source]↑

X-LINUX-AI Machine code.png
Only for STM32MP2 1
X-LINUX-AI TFLite.png
X-LINUX-AI Onnx.png
X-LINUX-AI coral.png
NPU/GPU CPU CPU EdgeTPU
Image classification logo.png
Image classification 		Yes Yes Yes Yes
Object detection logo.png
Object detection 		Yes Yes Yes Yes
Pose estimation logo.png
Pose estimation 		Yes No No No
Semantic segmentation logo.png
Semantic segmentation 		Yes No No No
Face recognition logo.png
Face recognition 		No Yes No No


Warning DB.png Important
1 NPU/GPU acceleration based on OpenVX backend using network binary graph (NBG) is available only for STM32MP2x boards

2. Performance Comparison[edit | edit source]↑


Model STM32MP25
NPU 		STM32MP15x
CPU
Image classification logo.png
Image classification 		MobilenetV1_0.5 128x128

515 FPS

22 FPS

MobilenetV2_1.0 224x224

72 FPS

2 FPS

Object detection logo.png
Object detection 		SSD_MobilenetV1_1.0 300x300

40 FPS

2 FPS

SSD_MobilenetV2_FPNLite 224x224 36 FPS 0.8 FPS
Face recognition logo.png
Face recognition 		STModel_FaceReco

15 FPS

5 FPS

Pose estimation logo.png
Pose estimation 		YoloV8n 256x256

59 FPS

1.3 FPS

Semantic segmentation logo.png
Semantic segmentation 		DeepLabV3 257x257

17 FPS

1 FPS



Microprocessor unit

Neural processing unit - NEW

Graphics processing units

Central processing unit

Frames per second

Retrieved from "https://wiki.st.com/stm32mpu-ecosystem-v5/index.php?title=X-LINUX-AI_application_samples_zoo&oldid=105374"