.jpg)
콘테스트 프로젝트 진행 상황
각 Quest를 단계적으로 완성하는 프로젝트입니다.
Quest 1: STM32Cube.AI 개발 환경 구축
2022.08.01
1. Analyzing report capture
2. Show Graph capture
3. report text
Neural Network Tools for STM32AI v1.6.0 (STM.ai v7.2.0-RC5)
Created date : 2022-07-21 14:07:49
Parameters : analyze --name network -m C:/Users/Leniven_HW/Downloads/STM32CubeFunctionPack_SENSING1_V4.0.3/Utilities/AI_Ressources/models/cnn_gmp.h5 --type keras --compression none --verbosity 1 --workspace C:\Users\LENIVE~1\AppData\Local\Temp\mxAI_workspace5105440905499006035669317230341452 --output C:\Users\Leniven_HW\.stm32cubemx\network_output --allocate-inputs --allocate-outputs
Exec/report summary (analyze)
------------------------------------------------------------------------------------------------------------------------
model file : C:\Users\Leniven_HW\Downloads\STM32CubeFunctionPack_SENSING1_V4.0.3\Utilities\AI_Ressources\models\cnn_gmp.h5
type : keras
c_name : network
compression : none
allocator strategy : ['allocate-inputs', 'allocate-outputs']
workspace dir : C:\Users\LENIVE~1\AppData\Local\Temp\mxAI_workspace5105440905499006035669317230341452
output dir : C:\Users\Leniven_HW\.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:activations/**default**
output 1/1 : 'softmax_1'
5 items, 20 B, ai_float, float, (1,1,1,5), domain:activations/**default**
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) : 6,976 B (6.81 KiB) = 6,976 + 0 + 0
(*) input/output buffers can be used from the activations buffer
Model name - cnn_gmp ['input_0'] ['softmax_1']
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
id layer (type,original) oshape param/size macc connected to | c_size c_macc c_type
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
0 input_0 (Input, None) [b:None,h:24,w:3,c:1] | +384(+100.0%) +5,776(+100.0%) conv2d_of32[0]
quantize_1_conv2d (Conv2D, Conv2D) [b:None,h:20,w:3,c:16] 96/384 4,816 input_0 | -384(-100.0%) -4,816(-100.0%)
quantize_1 (Nonlinearity, Conv2D) [b:None,h:20,w:3,c:16] 960 quantize_1_conv2d | -960(-100.0%)
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
1 quantize_2_conv2d (Conv2D, Conv2D) [b:None,h:16,w:3,c:16] 1,296/5,184 61,456 quantize_1 | -5,184(-100.0%) -61,456(-100.0%)
quantize_2 (Nonlinearity, Conv2D) [b:None,h:16,w:3,c:16] 768 quantize_2_conv2d | -768(-100.0%)
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
2 quantize_3 (Pool, GlobalMaxPooling2D) [b:None,h:1,w:1,c:16] 768 quantize_2 | +5,184(+100.0%) +62,224(+8102.1%) optimized_conv2d_of32[1]
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
3 quantize_4_dense (Dense, Dense) [b:None,h:1,w:1,c:5] 85/340 85 quantize_3 | dense_of32[2]
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
4 softmax_1 (Nonlinearity, Softmax) [b:None,h:1,w:1,c:5] 75 quantize_4_dense | nl_of32[o][3]
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
model/c-model: macc=68,928/68,928 weights=5,908/5,908 activations=--/6,976 io=--/0
Generated C-graph summary
------------------------------------------------------------------------------------------------------------------------
model name : cnn_gmp
c-name : network
c-node # : 4
c-array # : 12
activations size : 6976 (1 segments)
weights size : 5908 (1 segments)
macc : 68928
inputs : ['input_0_output']
outputs : ['softmax_1_output']
C-Arrays (12)
--------------------------------------------------------------------------------------------------------
c_id name (*_array) item/size domain/mem-pool c-type fmt comment
--------------------------------------------------------------------------------------------------------
0 input_0_output 72/288 activations/**default** float float /input
1 quantize_1_conv2d_output 960/3840 activations/**default** float float
2 quantize_2_conv2d_output 16/64 activations/**default** float float
3 quantize_4_dense_output 5/20 activations/**default** float float
4 softmax_1_output 5/20 activations/**default** float float /output
5 quantize_1_conv2d_weights 80/320 weights/weights const float float
6 quantize_1_conv2d_bias 16/64 weights/weights const float float
7 quantize_2_conv2d_weights 1280/5120 weights/weights const float float
8 quantize_2_conv2d_bias 16/64 weights/weights const float float
9 quantize_4_dense_weights 80/320 weights/weights const float float
10 quantize_4_dense_bias 5/20 weights/weights const float float
11 quantize_2_conv2d_scratch0 768/3072 activations/**default** float float
--------------------------------------------------------------------------------------------------------
C-Layers (4)
--------------------------------------------------------------------------------------------------------------------
c_id name (*_layer) id layer_type macc rom tensors shape (array id)
--------------------------------------------------------------------------------------------------------------------
0 quantize_1_conv2d 0 conv2d 5776 384 I: input_0_output (1,24,3,1) (0)
W: quantize_1_conv2d_weights (1,16,5,1) (5)
W: quantize_1_conv2d_bias (1,1,1,16) (6)
O: quantize_1_conv2d_output (1,20,3,16) (1)
--------------------------------------------------------------------------------------------------------------------
1 quantize_2_conv2d 2 optimized_conv2d 62992 5184 I: quantize_1_conv2d_output (1,20,3,16) (1)
S: quantize_2_conv2d_scratch0
W: quantize_2_conv2d_weights (16,16,5,1) (7)
W: quantize_2_conv2d_bias (1,1,1,16) (8)
O: quantize_2_conv2d_output (1,1,1,16) (2)
--------------------------------------------------------------------------------------------------------------------
2 quantize_4_dense 3 dense 85 340 I: quantize_2_conv2d_output (1,1,1,16) (2)
W: quantize_4_dense_weights (16,1,1,5) (9)
W: quantize_4_dense_bias (1,1,1,5) (10)
O: quantize_4_dense_output (1,1,1,5) (3)
--------------------------------------------------------------------------------------------------------------------
3 softmax_1 4 nl 75 0 I: quantize_4_dense_output (1,1,1,5) (3)
O: softmax_1_output (1,1,1,5) (4)
--------------------------------------------------------------------------------------------------------------------
Number of operations per c-layer
---------------------------------------------------------------------------------------------------------------
c_id m_id name (type) #op (type) #param (sparsity)
---------------------------------------------------------------------------------------------------------------
0 0 quantize_1_conv2d (conv2d) 5,776 (smul_f32_f32) 96 (0.0000)
1 2 quantize_2_conv2d (optimized_conv2d) 62,992 (smul_f32_f32) 1,296 (0.0000)
2 3 quantize_4_dense (dense) 85 (smul_f32_f32) 85 (0.1176)
3 4 softmax_1 (nl) 75 (op_f32_f32)
---------------------------------------------------------------------------------------------------------------
total 68,928 1,477 (0.0068)
Number of operation types
---------------------------------------------
smul_f32_f32 68,853 99.9%
op_f32_f32 75 0.1%
Complexity report (model)
-----------------------------------------------------------------------------------
m_id name c_macc c_rom c_id
-----------------------------------------------------------------------------------
0 input_0 || 8.4% || 6.5% [0]
2 quantize_3 |||||||||||||||| 91.4% |||||||||||||||| 87.7% [1]
3 quantize_4_dense | 0.1% | 5.8% [2]
4 softmax_1 | 0.1% | 0.0% [3]
-----------------------------------------------------------------------------------
macc=68,928 weights=5,908 act=6,976 ram_io=0
Quest 2: Neural Network Model 생성 개발 환경 구축하기
2022.08.09
1. Accuracy 캡처
2. Model Loss Function 캡처
3. Confusion Matrix 캡처
4. netwrok_analyze_report
Neural Network Tools for STM32AI v1.6.0 (STM.ai v7.1.0-RC3)
Created date : 2022-08-10 11:12:20
Parameters : analyze --name network -m C:/Users/Leniven_HW/PycharmProjects/STM32_Quest_AI_GIGECAP_2022/HAR/results/2022_Aug_10_10_27_05/har_IGN.h5 --type keras --compression 1 --verbosity 1 --workspace C:\Users\LENIVE~1\AppData\Local\Temp\mxAI_workspace15824329646230011944412612009690500 --output C:\Users\Leniven_HW\.stm32cubemx\network_output
Exec/report summary (analyze)
------------------------------------------------------------------------------------------------------------------------
model file : C:\Users\Leniven_HW\PycharmProjects\STM32_Quest_AI_GIGECAP_2022\HAR\results\2022_Aug_10_10_27_05\har_IGN.h5
type : keras
c_name : network
compression : None
workspace dir : C:\Users\LENIVE~1\AppData\Local\Temp\mxAI_workspace15824329646230011944412612009690500
output dir : C:\Users\Leniven_HW\.stm32cubemx\network_output
model_name : har_IGN
model_hash : 85e072a41dbcb6be927255cf329473ae
input 1/1 : 'input_0'
72 items, 288 B, ai_float, float, (1,24,3,1), domain:user/
output 1/1 : 'dense_1'
4 items, 16 B, ai_float, float, (1,1,1,4), domain:user/
params # : 3,064 items (11.97 KiB)
macc : 14,404
weights (ro) : 12,256 B (11.97 KiB) (1 segment)
activations (rw) : 1,728 B (1.69 KiB) (1 segment)
ram (total) : 2,032 B (1.98 KiB) = 1,728 + 288 + 16
Model name - har_IGN ['input_0'] ['dense_1']
-------------------------------------------------------------------------------------------------------------------------------------------
id layer (type) oshape param/size macc connected to | c_size c_macc c_type
-------------------------------------------------------------------------------------------------------------------------------------------
0 input_0 (Input) (None,24,3,1) |
conv2d_conv2d (Conv2D) (None,9,3,24) 408/1,632 10,392 input_0 | +1,296(+12.5%) optimized_conv2d()[0]
conv2d (Nonlinearity) (None,9,3,24) 648 conv2d_conv2d | -648(-100.0%)
-------------------------------------------------------------------------------------------------------------------------------------------
1 max_pooling2d (Pool) (None,3,3,24) 648 conv2d | -648(-100.0%)
-------------------------------------------------------------------------------------------------------------------------------------------
2 flatten (Reshape) (None,1,1,216) max_pooling2d |
-------------------------------------------------------------------------------------------------------------------------------------------
3 dense_dense (Dense) (None,1,1,12) 2,604/10,416 2,604 flatten | dense()[1]
-------------------------------------------------------------------------------------------------------------------------------------------
5 dense_1_dense (Dense) (None,1,1,4) 52/208 52 dense_dense | dense()[2]
dense_1 (Nonlinearity) (None,1,1,4) 60 dense_1_dense | nl()/o[3]
-------------------------------------------------------------------------------------------------------------------------------------------
model/c-model: macc=14,404/14,404 weights=12,256/12,256 activations=--/1,728 io=--/304
Generated C-graph summary
------------------------------------------------------------------------------------------------------------------------
model name : har_ign
c-name : network
c-node # : 4
c-array # : 12
activations size : 1728 (1 segments)
weights size : 12256 (1 segments)
macc : 14404
inputs : ['input_0_output']
outputs : ['dense_1_output']
C-Arrays (12)
-----------------------------------------------------------------------------------------------------
c_id name (*_array) item/size domain/mem-pool c-type fmt comment
-----------------------------------------------------------------------------------------------------
0 input_0_output 72/288 user/ float float /input
1 conv2d_conv2d_output 216/864 activations/**default** float float
2 dense_dense_output 12/48 activations/**default** float float
3 dense_1_dense_output 4/16 activations/**default** float float
4 dense_1_output 4/16 user/ float float /output
5 conv2d_conv2d_weights 384/1536 weights/ const float float
6 conv2d_conv2d_bias 24/96 weights/ const float float
7 dense_dense_weights 2592/10368 weights/ const float float
8 dense_dense_bias 12/48 weights/ const float float
9 dense_1_dense_weights 48/192 weights/ const float float
10 dense_1_dense_bias 4/16 weights/ const float float
11 conv2d_conv2d_scratch0 216/864 activations/**default** float float
-----------------------------------------------------------------------------------------------------
C-Layers (4)
-------------------------------------------------------------------------------------------------------------
c_id name (*_layer) id layer_type macc rom tensors shape (array id)
-------------------------------------------------------------------------------------------------------------
0 conv2d_conv2d 1 optimized_conv2d 11688 1632 I: input_0_output (1,24,3,1) (0)
S: conv2d_conv2d_scratch0
W: conv2d_conv2d_weights (1,24,16,1) (5)
W: conv2d_conv2d_bias (1,1,1,24) (6)
O: conv2d_conv2d_output (1,3,3,24) (1)
-------------------------------------------------------------------------------------------------------------
1 dense_dense 3 dense 2604 10416 I: conv2d_conv2d_output0 (1,1,1,216) (1)
W: dense_dense_weights (216,1,1,12) (7)
W: dense_dense_bias (1,1,1,12) (8)
O: dense_dense_output (1,1,1,12) (2)
-------------------------------------------------------------------------------------------------------------
2 dense_1_dense 5 dense 52 208 I: dense_dense_output (1,1,1,12) (2)
W: dense_1_dense_weights (12,1,1,4) (9)
W: dense_1_dense_bias (1,1,1,4) (10)
O: dense_1_dense_output (1,1,1,4) (3)
-------------------------------------------------------------------------------------------------------------
3 dense_1 5 nl 60 0 I: dense_1_dense_output (1,1,1,4) (3)
O: dense_1_output (1,1,1,4) (4)
-------------------------------------------------------------------------------------------------------------
Complexity report per layer - macc=14,404 weights=12,256 act=1,728 ram_io=304
------------------------------------------------------------------------------
id name c_macc c_rom c_id
------------------------------------------------------------------------------
0 conv2d_conv2d |||||||||||||||| 81.1% ||| 13.3% [0]
3 dense_dense |||| 18.1% |||||||||||||||| 85.0% [1]
5 dense_1_dense | 0.4% | 1.7% [2]
5 dense_1 | 0.4% | 0.0% [3]
프로젝트 갤러리
프로젝트 제작 과정과 최종 결과물을 담은 이미지 갤러리입니다.
프로젝트 자료실
프로젝트를 재현하거나 학습하는데 필요한 모든 자료를 제공합니다.
등록된 자료가 없습니다.
프로젝트 소개
아직 등록된 프로젝트 소개가 없습니다.
코드 & 회로도
등록된 코드/회로 자료가 없습니다.
댓글 0개
아직 댓글이 없습니다. 첫 댓글을 남겨보세요!