수행기록퀘스트1

개발 환경 구축
2022. 8. 5 (금) 00:09 Milli 조회 396 좋아요 0 스크랩 0 댓글 0

1. MCU 선정

STM32L475VGTx MCU를 선택

 

2. 소프트웨어 팩 선택

-X cube AI 설정

 

3.MCU 설정 및  소프트웨어 팩 설정

4. Analyuze 시료

Analyzing model 
C:/Users/kjh14/STM32Cube/Repository/Packs/STMicroelectronics/X-CUBE-AI/7.1.0/Utilities/windows/stm32ai analyze --name network -m F:/Util/Software tool/STM/en.fp-ai-sensing1/STM32CubeFunctionPack_SENSING1_V4.0.3/Utilities/AI_Ressources/models/cnn_gmp.h5 --type keras --compression 1 --verbosity 1 --workspace C:\Users\kjh14\AppData\Local\Temp\mxAI_workspace169198008605006123145192526529961 --output C:\Users\kjh14\.stm32cubemx\network_output  
Neural Network Tools for STM32AI v1.6.0 (STM.ai v7.1.0-RC3) 
  
 Exec/report summary (analyze) 
 ------------------------------------------------------------------------------------------------------------------------ 
 model file           : F:\Util\Software tool\STM\en.fp-ai-sensing1\STM32CubeFunctionPack_SENSING1_V4.0.3\Utilities\AI_Ressources\models\cnn_gmp.h5 
 type                 : keras 
 c_name               : network 
 compression          : None 
 workspace dir        : C:\Users\kjh14\AppData\Local\Temp\mxAI_workspace169198008605006123145192526529961 
 output dir           : C:\Users\kjh14\.stm32cubemx\network_output 
 model_name           : cnn_gmp 
 model_hash           : 954164d0a35496bd293bb6b3429c6a79 
 input 1/1            : 'input_0' 
                        72 items, 288 B, ai_float, float, (1,24,3,1), domain:user/ 
 output 1/1           : 'softmax_1' 
                        5 items, 20 B, ai_float, float, (1,1,1,5), domain:user/ 
 params #             : 1,477 items (5.77 KiB) 
 macc                 : 68,928 
 weights (ro)         : 5,908 B (5.77 KiB) (1 segment) 
 activations (rw)     : 6,976 B (6.81 KiB) (1 segment) 
 ram (total)          : 7,284 B (7.11 KiB) = 6,976 + 288 + 20 
  
 Model name - cnn_gmp ['input_0'] ['softmax_1'] 
 ------------------------------------------------------------------------------------------------------------------------------------------------- 
 id   layer (type)                 oshape           param/size    macc     connected to        |   c_size   c_macc          c_type                
 ------------------------------------------------------------------------------------------------------------------------------------------------- 
 0    input_0 (Input)              (None,24,3,1)                                               |                            
      quantize_1_conv2d (Conv2D)   (None,20,3,16)   96/384        4,816    input_0             |            +960(+19.9%)    conv2d()[0]           
      quantize_1 (Nonlinearity)    (None,20,3,16)                 960      quantize_1_conv2d   |            -960(-100.0%)   
 ------------------------------------------------------------------------------------------------------------------------------------------------- 
 1    quantize_2_conv2d (Conv2D)   (None,16,3,16)   1,296/5,184   61,456   quantize_1          |            +1,536(+2.5%)   optimized_conv2d()[1] 
      quantize_2 (Nonlinearity)    (None,16,3,16)                 768      quantize_2_conv2d   |            -768(-100.0%)   
 ------------------------------------------------------------------------------------------------------------------------------------------------- 
 2    quantize_3 (Pool)            (None,1,1,16)                  768      quantize_2          |            -768(-100.0%)   
 ------------------------------------------------------------------------------------------------------------------------------------------------- 
 3    quantize_4_dense (Dense)     (None,1,1,5)     85/340        85       quantize_3          |                            dense()[2]            
 ------------------------------------------------------------------------------------------------------------------------------------------------- 
 4    softmax_1 (Nonlinearity)     (None,1,1,5)                   75       quantize_4_dense    |                            nl()/o[3]             
 ------------------------------------------------------------------------------------------------------------------------------------------------- 
 model/c-model: macc=68,928/68,928  weights=5,908/5,908  activations=--/6,976 io=--/308 
  
 Complexity report per layer - macc=68,928 weights=5,908 act=6,976 ram_io=308 
 ---------------------------------------------------------------------------------- 
 id   name                c_macc                    c_rom                     c_id 
 ---------------------------------------------------------------------------------- 
 0    quantize_1_conv2d   ||                 8.4%   ||                 6.5%   [0]  
 1    quantize_2_conv2d   ||||||||||||||||  91.4%   ||||||||||||||||  87.7%   [1]  
 3    quantize_4_dense    |                  0.1%   |                  5.8%   [2]  
 4    softmax_1           |                  0.1%   |                  0.0%   [3] 
Creating txt report file C:\Users\kjh14\.stm32cubemx\network_output\network_analyze_report.txt 
elapsed time (analyze): 0.577s 
Analyze complete on AI model

5. Network

개발환경 구축 완료

첨부파일
STM32cubeMX.AI.zip 다운로드

로그인 후
참가 상태를 확인할 수 있습니다.