内容介绍
内容介绍
1.项目介绍
项目基于官方的示例工程Snake Game, 通过语音识别“up,down,left,right...”等关键词，来控制蛇的移动，硬件采用MAX78000FTHR开发板，集成基于硬件的卷积神经网络加速器可以执行 AI 推理在非常低的能量水平，显示屏采用DFROBOT生产的1.54寸LCD显示模块。
2.项目设计思路
2.1.搭建环境
模型训练的环境搭建参照官方说明文档，实际搭建过程中，根据硬件环境，以及系统版本会有各种不同问题。
本项目采用系统为Ubuntu20.04 lts主要搭建步骤如下：
--安装项目所需的工具包，注意斜杠某些系统环境无法识别斜杠\需要分多次安装
$ sudo apt-get install -y make build-essential libssl-dev zlib1g-dev \
libbz2-dev libreadline-dev libsqlite3-dev wget curl llvm \
libncurses5-dev libncursesw5-dev xz-utils tk-dev libffi-dev liblzma-dev \
libsndfile-dev portaudio19-dev--确认系统python版本为Python 3.8.x
--克隆项目文件
$ git clone --recursive https://github.com/MaximIntegratedAI/ai8x-training.git
$ git clone --recursive https://github.com/MaximIntegratedAI/ai8x-synthesis.git--安装项目依赖包
(ai8x-training) $ pip3 install -U pip wheel setuptools
(ai8x-training) $ pip3 install -r requirements-cu11.txt2.2 执行训练
./scripts/train_kws20_v3.sh2.3.环境搭建过程中遇到的各种问题
ERROR: resampy 0.4.2 has requirement numba>=0.53, but you'll have numba 0.49.1 which is incompatible.
ERROR: tensorboard 2.9.1 has requirement protobuf<3.20,>=3.9.2, but you'll have protobuf 3.20.3 which is incompatible.
ERROR: nbconvert 7.2.5 has requirement jinja2>=3.0, but you'll have jinja2 2.10.1 which is incompatible.
ERROR: nbconvert 7.2.5 has requirement pygments>=2.4.1, but you'll have pygments 2.3.1 which is incompatible.
ERROR: ipython 8.7.0 has requirement pygments>=2.4.0, but you'll have pygments 2.3.1 which is incompatible.按照错误提示手动安装各个版本包
$ pip install numba==0.53
$ pip install protobuf==3.9.2
$ pip install jinja2==3.0
$ pip install pygments==2.4.1
3.搜集素材的思路
由于项目目的为识别用户自定义关键字，因此需要采集用户关键字语音信息，为保证模型效果，需要分多次，各种环境背景音下，采集用户关键字，采集到的语音整理为以下格式。
时间为1s，32kB，16-bit小端 PCM编码 WAVE，采样率16k
4.预训练实现过程
经过了几周的尝试，环境终于搭建好了 ，从最初的wsl2，到腾讯云，阿里云各个环境都试了一遍，同样是一套安装流程，报的错是各种各样的，最终还是回到wsl2上面，好在环境是终于搭建好了，训练也跑起来了，由于用的cpu平台，训练速度有些慢。
Training epoch: 258930 samples (256 per mini-batch)
Epoch: [19][ 10/ 1012] Overall Loss 0.313984 Objective Loss 0.313984 LR 0.001000 Time 0.287783
Epoch: [19][ 20/ 1012] Overall Loss 0.301635 Objective Loss 0.301635 LR 0.001000 Time 0.199325
Epoch: [19][ 30/ 1012] Overall Loss 0.294099 Objective Loss 0.294099 LR 0.001000 Time 0.169220
Epoch: [19][ 40/ 1012] Overall Loss 0.292085 Objective Loss 0.292085 LR 0.001000 Time 0.153846
Epoch: [19][ 50/ 1012] Overall Loss 0.289668 Objective Loss 0.289668 LR 0.001000 Time 0.144684
Epoch: [19][ 60/ 1012] Overall Loss 0.286566 Objective Loss 0.286566 LR 0.001000 Time 0.138351
Epoch: [19][ 70/ 1012] Overall Loss 0.288767 Objective Loss 0.288767 LR 0.001000 Time 0.134098
Epoch: [19][ 80/ 1012] Overall Loss 0.293444 Objective Loss 0.293444 LR 0.001000 Time 0.130837
Epoch: [19][ 90/ 1012] Overall Loss 0.295964 Objective Loss 0.295964 LR 0.001000 Time 0.128300
Epoch: [19][ 100/ 1012] Overall Loss 0.294188 Objective Loss 0.294188 LR 0.001000 Time 0.126260
Epoch: [19][ 110/ 1012] Overall Loss 0.296818 Objective Loss 0.296818 LR 0.001000 Time 0.124561
Epoch: [19][ 120/ 1012] Overall Loss 0.294679 Objective Loss 0.294679 LR 0.001000 Time 0.123139
Epoch: [19][ 130/ 1012] Overall Loss 0.295100 Objective Loss 0.295100 LR 0.001000 Time 0.121976
Epoch: [19][ 140/ 1012] Overall Loss 0.295078 Objective Loss 0.295078 LR 0.001000 Time 0.121083
Epoch: [19][ 150/ 1012] Overall Loss 0.293299 Objective Loss 0.293299 LR 0.001000 Time 0.120260
Epoch: [19][ 160/ 1012] Overall Loss 0.293785 Objective Loss 0.293785 LR 0.001000 Time 0.119530
Epoch: [19][ 170/ 1012] Overall Loss 0.293703 Objective Loss 0.293703 LR 0.001000 Time 0.118871
Epoch: [19][ 180/ 1012] Overall Loss 0.292900 Objective Loss 0.292900 LR 0.001000 Time 0.118241
Epoch: [19][ 190/ 1012] Overall Loss 0.292962 Objective Loss 0.292962 LR 0.001000 Time 0.117733
Epoch: [19][ 200/ 1012] Overall Loss 0.293803 Objective Loss 0.293803 LR 0.001000 Time 0.117261
Epoch: [19][ 210/ 1012] Overall Loss 0.294633 Objective Loss 0.294633 LR 0.001000 Time 0.116829
Epoch: [19][ 220/ 1012] Overall Loss 0.294978 Objective Loss 0.294978 LR 0.001000 Time 0.116448
Epoch: [19][ 230/ 1012] Overall Loss 0.294477 Objective Loss 0.294477 LR 0.001000 Time 0.116269
Epoch: [19][ 240/ 1012] Overall Loss 0.293564 Objective Loss 0.293564 LR 0.001000 Time 0.116011
Epoch: [19][ 250/ 1012] Overall Loss 0.292999 Objective Loss 0.292999 LR 0.001000 Time 0.115683
Epoch: [19][ 260/ 1012] Overall Loss 0.293600 Objective Loss 0.293600 LR 0.001000 Time 0.115365
Epoch: [19][ 270/ 1012] Overall Loss 0.293604 Objective Loss 0.293604 LR 0.001000 Time 0.115125
Epoch: [19][ 280/ 1012] Overall Loss 0.295085 Objective Loss 0.295085 LR 0.001000 Time 0.114850
Epoch: [19][ 290/ 1012] Overall Loss 0.296082 Objective Loss 0.296082 LR 0.001000 Time 0.114626
Epoch: [19][ 300/ 1012] Overall Loss 0.296814 Objective Loss 0.296814 LR 0.001000 Time 0.114418
Epoch: [19][ 310/ 1012] Overall Loss 0.297363 Objective Loss 0.297363 LR 0.001000 Time 0.114225
Epoch: [19][ 320/ 1012] Overall Loss 0.297183 Objective Loss 0.297183 LR 0.001000 Time 0.114022
Epoch: [19][ 330/ 1012] Overall Loss 0.297062 Objective Loss 0.297062 LR 0.001000 Time 0.113859
Epoch: [19][ 340/ 1012] Overall Loss 0.297388 Objective Loss 0.297388 LR 0.001000 Time 0.113696
Epoch: [19][ 350/ 1012] Overall Loss 0.296915 Objective Loss 0.296915 LR 0.001000 Time 0.113558
Epoch: [19][ 360/ 1012] Overall Loss 0.296891 Objective Loss 0.296891 LR 0.001000 Time 0.113428
Epoch: [19][ 370/ 1012] Overall Loss 0.296052 Objective Loss 0.296052 LR 0.001000 Time 0.113292
Epoch: [19][ 380/ 1012] Overall Loss 0.294866 Objective Loss 0.294866 LR 0.001000 Time 0.113174
Epoch: [19][ 390/ 1012] Overall Loss 0.294852 Objective Loss 0.294852 LR 0.001000 Time 0.113037
Epoch: [19][ 400/ 1012] Overall Loss 0.294815 Objective Loss 0.294815 LR 0.001000 Time 0.112923
Epoch: [19][ 410/ 1012] Overall Loss 0.295323 Objective Loss 0.295323 LR 0.001000 Time 0.112826
Epoch: [19][ 420/ 1012] Overall Loss 0.294929 Objective Loss 0.294929 LR 0.001000 Time 0.112732
Epoch: [19][ 430/ 1012] Overall Loss 0.295521 Objective Loss 0.295521 LR 0.001000 Time 0.112623
Epoch: [19][ 440/ 1012] Overall Loss 0.295394 Objective Loss 0.295394 LR 0.001000 Time 0.112523
Epoch: [19][ 450/ 1012] Overall Loss 0.295120 Objective Loss 0.295120 LR 0.001000 Time 0.112425
Epoch: [19][ 460/ 1012] Overall Loss 0.295059 Objective Loss 0.295059 LR 0.001000 Time 0.112304
Epoch: [19][ 470/ 1012] Overall Loss 0.295690 Objective Loss 0.295690 LR 0.001000 Time 0.112206
Epoch: [19][ 480/ 1012] Overall Loss 0.295448 Objective Loss 0.295448 LR 0.001000 Time 0.112127
Epoch: [19][ 490/ 1012] Overall Loss 0.295219 Objective Loss 0.295219 LR 0.001000 Time 0.112067
Epoch: [19][ 500/ 1012] Overall Loss 0.294877 Objective Loss 0.294877 LR 0.001000 Time 0.111992
Epoch: [19][ 510/ 1012] Overall Loss 0.294220 Objective Loss 0.294220 LR 0.001000 Time 0.111924
Epoch: [19][ 520/ 1012] Overall Loss 0.293924 Objective Loss 0.293924 LR 0.001000 Time 0.111864
Epoch: [19][ 530/ 1012] Overall Loss 0.294121 Objective Loss 0.294121 LR 0.001000 Time 0.111797
Epoch: [19][ 540/ 1012] Overall Loss 0.294086 Objective Loss 0.294086 LR 0.001000 Time 0.111725
Epoch: [19][ 550/ 1012] Overall Loss 0.294411 Objective Loss 0.294411 LR 0.001000 Time 0.111668
Epoch: [19][ 560/ 1012] Overall Loss 0.294173 Objective Loss 0.294173 LR 0.001000 Time 0.111604
Epoch: [19][ 570/ 1012] Overall Loss 0.294153 Objective Loss 0.294153 LR 0.001000 Time 0.111563
Epoch: [19][ 580/ 1012] Overall Loss 0.294451 Objective Loss 0.294451 LR 0.001000 Time 0.111511
Epoch: [19][ 590/ 1012] Overall Loss 0.294247 Objective Loss 0.294247 LR 0.001000 Time 0.111468
Epoch: [19][ 600/ 1012] Overall Loss 0.294086 Objective Loss 0.294086 LR 0.001000 Time 0.111423
Epoch: [19][ 610/ 1012] Overall Loss 0.294067 Objective Loss 0.294067 LR 0.001000 Time 0.111381
Epoch: [19][ 620/ 1012] Overall Loss 0.294201 Objective Loss 0.294201 LR 0.001000 Time 0.111335
Epoch: [19][ 630/ 1012] Overall Loss 0.293960 Objective Loss 0.293960 LR 0.001000 Time 0.111267
Epoch: [19][ 640/ 1012] Overall Loss 0.294124 Objective Loss 0.294124 LR 0.001000 Time 0.111218
Epoch: [19][ 650/ 1012] Overall Loss 0.293984 Objective Loss 0.293984 LR 0.001000 Time 0.111162
Epoch: [19][ 660/ 1012] Overall Loss 0.293730 Objective Loss 0.293730 LR 0.001000 Time 0.111119
Epoch: [19][ 670/ 1012] Overall Loss 0.293675 Objective Loss 0.293675 LR 0.001000 Time 0.111083
Epoch: [19][ 680/ 1012] Overall Loss 0.294043 Objective Loss 0.294043 LR 0.001000 Time 0.111051
Epoch: [19][ 690/ 1012] Overall Loss 0.294001 Objective Loss 0.294001 LR 0.001000 Time 0.111014
Epoch: [19][ 700/ 1012] Overall Loss 0.293943 Objective Loss 0.293943 LR 0.001000 Time 0.110990
Epoch: [19][ 710/ 1012] Overall Loss 0.294046 Objective Loss 0.294046 LR 0.001000 Time 0.110950
Epoch: [19][ 720/ 1012] Overall Loss 0.294187 Objective Loss 0.294187 LR 0.001000 Time 0.110913
Epoch: [19][ 730/ 1012] Overall Loss 0.294282 Objective Loss 0.294282 LR 0.001000 Time 0.110876
Epoch: [19][ 740/ 1012] Overall Loss 0.294543 Objective Loss 0.294543 LR 0.001000 Time 0.110850
Epoch: [19][ 750/ 1012] Overall Loss 0.294685 Objective Loss 0.294685 LR 0.001000 Time 0.110810
Epoch: [19][ 760/ 1012] Overall Loss 0.294658 Objective Loss 0.294658 LR 0.001000 Time 0.110775
Epoch: [19][ 770/ 1012] Overall Loss 0.294649 Objective Loss 0.294649 LR 0.001000 Time 0.110737
Epoch: [19][ 780/ 1012] Overall Loss 0.294585 Objective Loss 0.294585 LR 0.001000 Time 0.110692
Epoch: [19][ 790/ 1012] Overall Loss 0.294528 Objective Loss 0.294528 LR 0.001000 Time 0.110665
Epoch: [19][ 800/ 1012] Overall Loss 0.294613 Objective Loss 0.294613 LR 0.001000 Time 0.110661
Epoch: [19][ 810/ 1012] Overall Loss 0.294572 Objective Loss 0.294572 LR 0.001000 Time 0.110618
Epoch: [19][ 820/ 1012] Overall Loss 0.294865 Objective Loss 0.294865 LR 0.001000 Time 0.110589
Epoch: [19][ 830/ 1012] Overall Loss 0.294934 Objective Loss 0.294934 LR 0.001000 Time 0.110562
Epoch: [19][ 840/ 1012] Overall Loss 0.294989 Objective Loss 0.294989 LR 0.001000 Time 0.110527
Epoch: [19][ 850/ 1012] Overall Loss 0.295124 Objective Loss 0.295124 LR 0.001000 Time 0.110504
Epoch: [19][ 860/ 1012] Overall Loss 0.295317 Objective Loss 0.295317 LR 0.001000 Time 0.110476
Epoch: [19][ 870/ 1012] Overall Loss 0.295224 Objective Loss 0.295224 LR 0.001000 Time 0.110447
Epoch: [19][ 880/ 1012] Overall Loss 0.295169 Objective Loss 0.295169 LR 0.001000 Time 0.110420
Epoch: [19][ 890/ 1012] Overall Loss 0.295155 Objective Loss 0.295155 LR 0.001000 Time 0.110394
Epoch: [19][ 900/ 1012] Overall Loss 0.295399 Objective Loss 0.295399 LR 0.001000 Time 0.110379
Epoch: [19][ 910/ 1012] Overall Loss 0.295536 Objective Loss 0.295536 LR 0.001000 Time 0.110358
Epoch: [19][ 920/ 1012] Overall Loss 0.295419 Objective Loss 0.295419 LR 0.001000 Time 0.110338
Epoch: [19][ 930/ 1012] Overall Loss 0.295352 Objective Loss 0.295352 LR 0.001000 Time 0.110310
Epoch: [19][ 940/ 1012] Overall Loss 0.295218 Objective Loss 0.295218 LR 0.001000 Time 0.110284
Epoch: [19][ 950/ 1012] Overall Loss 0.295563 Objective Loss 0.295563 LR 0.001000 Time 0.110261
Epoch: [19][ 960/ 1012] Overall Loss 0.295651 Objective Loss 0.295651 LR 0.001000 Time 0.110249
Epoch: [19][ 970/ 1012] Overall Loss 0.295934 Objective Loss 0.295934 LR 0.001000 Time 0.110222
Epoch: [19][ 980/ 1012] Overall Loss 0.296039 Objective Loss 0.296039 LR 0.001000 Time 0.110215
Epoch: [19][ 990/ 1012] Overall Loss 0.295852 Objective Loss 0.295852 LR 0.001000 Time 0.110195
Epoch: [19][ 1000/ 1012] Overall Loss 0.295951 Objective Loss 0.295951 LR 0.001000 Time 0.110185
Epoch: [19][ 1010/ 1012] Overall Loss 0.296195 Objective Loss 0.296195 LR 0.001000 Time 0.110187
Epoch: [19][ 1012/ 1012] Overall Loss 0.296139 Objective Loss 0.296139 Top1 85.135135 Top5 98.108108 LR 0.001000 Time 0.110134
--- validate (epoch=19)-----------
28770 samples (256 per mini-batch)
Epoch: [19][ 10/ 113] Loss 0.326720 Top1 86.250000 Top5 98.476562
Epoch: [19][ 20/ 113] Loss 0.320755 Top1 85.507812 Top5 98.437500
Epoch: [19][ 30/ 113] Loss 0.324655 Top1 85.625000 Top5 98.307292
Epoch: [19][ 40/ 113] Loss 0.324007 Top1 85.615234 Top5 98.271484
Epoch: [19][ 50/ 113] Loss 0.321523 Top1 85.757812 Top5 98.281250
Epoch: [19][ 60/ 113] Loss 0.325594 Top1 85.390625 Top5 98.229167
Epoch: [19][ 70/ 113] Loss 0.328396 Top1 85.267857 Top5 98.208705
Epoch: [19][ 80/ 113] Loss 0.325646 Top1 85.385742 Top5 98.271484
Epoch: [19][ 90/ 113] Loss 0.325350 Top1 85.447049 Top5 98.272569
Epoch: [19][ 100/ 113] Loss 0.327869 Top1 85.347656 Top5 98.281250
Epoch: [19][ 110/ 113] Loss 0.327079 Top1 85.394176 Top5 98.306108
Epoch: [19][ 113/ 113] Loss 0.327506 Top1 85.345846 Top5 98.300313
==> Top1: 85.346 Top5: 98.300 Loss: 0.328
==> Confusion:
[[ 856 0 1 1 13 3 0 0 4 101 0 5 0 1 2 2 0 2 0 3 17]
[ 4 973 1 4 5 21 3 11 0 2 2 1 0 0 3 4 1 2 7 5 12]
[ 3 2 963 10 2 1 22 6 1 2 2 3 2 6 3 5 2 1 5 5 11]
[ 3 1 18 935 1 2 3 1 0 0 6 1 2 1 27 1 1 3 12 1 11]
[ 12 11 3 4 932 6 0 1 2 6 0 3 0 2 6 3 7 0 0 3 12]
[ 1 49 1 8 5 915 3 39 2 2 1 10 4 7 2 1 0 1 2 5 14]
[ 2 2 19 0 0 2 1046 2 0 0 0 1 2 0 1 8 0 2 0 11 7]
[ 1 15 11 2 0 24 7 922 1 1 0 3 1 0 0 2 0 0 28 21 7]
[ 9 4 0 2 1 3 0 2 941 42 6 1 0 14 13 3 2 0 0 0 5]
[ 38 0 1 0 4 2 1 0 27 923 1 2 0 21 5 3 0 1 0 2 8]
[ 4 4 5 5 1 2 4 5 26 1 1009 0 2 11 3 1 2 1 14 4 15]
[ 2 1 0 0 0 11 2 4 0 0 1 924 20 5 0 11 4 4 0 22 13]
[ 1 0 2 7 2 1 0 2 2 0 0 47 905 1 3 6 5 22 1 5 41]
[ 1 0 3 0 6 15 0 1 10 9 4 8 1 916 5 4 3 0 0 8 15]
[ 4 3 1 19 10 1 1 0 26 4 2 0 2 1 944 0 3 2 8 2 16]
[ 3 2 3 2 4 1 4 1 1 0 0 6 2 2 1 982 6 7 0 3 8]
[ 1 6 2 2 11 2 0 0 1 3 1 3 1 1 3 11 997 1 0 4 30]
[ 6 1 1 9 0 1 3 1 1 0 0 9 21 5 0 19 1 969 0 4 8]
[ 2 4 5 21 1 0 0 16 2 0 6 1 1 0 15 1 1 0 937 2 10]
[ 0 2 0 0 1 2 5 9 0 0 0 12 3 4 0 5 8 3 0 1011 6]
[ 114 152 124 64 84 107 43 118 80 91 63 94 206 243 100 84 88 28 127 197 5554]]
==> Best [Top1: 85.666 Top5: 98.377 Sparsity:0.00 Params: 169472 on epoch: 16]
Saving checkpoint to: logs/2023.01.28-232929/qat_checkpoint.pth.tar
--- test ---------------------
29787 samples (256 per mini-batch)
Test: [ 10/ 117] Loss 0.326919 Top1 85.390625 Top5 98.476562
Test: [ 20/ 117] Loss 0.337157 Top1 85.312500 Top5 98.378906
Test: [ 30/ 117] Loss 0.336248 Top1 85.403646 Top5 98.385417
Test: [ 40/ 117] Loss 0.341029 Top1 85.175781 Top5 98.339844
Test: [ 50/ 117] Loss 0.343385 Top1 84.960938 Top5 98.343750
Test: [ 60/ 117] Loss 0.347126 Top1 84.960938 Top5 98.365885
Test: [ 70/ 117] Loss 0.347996 Top1 84.877232 Top5 98.342634
Test: [ 80/ 117] Loss 0.348859 Top1 84.882812 Top5 98.300781
Test: [ 90/ 117] Loss 0.350195 Top1 84.791667 Top5 98.355035
Test: [ 100/ 117] Loss 0.350307 Top1 84.800781 Top5 98.351562
Test: [ 110/ 117] Loss 0.348503 Top1 84.747869 Top5 98.338068
Test: [ 117/ 117] Loss 0.350601 Top1 84.728237 Top5 98.324773
==> Top1: 84.728 Top5: 98.325 Loss: 0.351
==> Confusion:
[[ 874 1 4 1 14 3 1 1 10 121 2 0 1 4 4 3 2 1 1 0 11]
[ 2 1023 3 3 8 23 1 16 1 1 0 0 0 3 2 3 3 2 11 7 19]
[ 6 2 998 20 3 1 26 12 0 4 2 0 0 1 3 4 1 1 6 3 8]
[ 5 0 21 968 0 3 3 0 0 0 13 0 1 1 28 2 2 4 13 3 7]
[ 13 11 1 0 976 8 0 0 0 17 0 1 1 2 4 2 9 0 0 2 12]
[ 1 43 1 6 6 946 2 36 2 1 1 8 4 16 1 0 3 2 5 5 15]
[ 0 0 24 1 0 2 1086 3 0 0 1 0 0 1 1 7 0 0 0 6 5]
[ 1 15 9 1 5 36 3 1030 1 1 0 4 0 2 1 1 0 1 34 9 13]
[ 4 4 1 1 0 0 0 0 922 37 6 0 1 11 19 1 3 1 6 0 12]
[ 34 0 1 3 4 2 0 3 39 934 0 0 1 23 8 4 1 0 1 3 13]
[ 2 3 6 11 0 2 4 2 23 2 952 0 1 15 8 1 3 0 20 6 16]
[ 2 1 4 0 0 19 4 5 1 0 0 986 22 5 3 5 3 8 1 24 14]
[ 5 1 4 8 1 2 3 4 2 0 4 74 943 0 3 13 11 16 3 10 30]
[ 0 1 1 0 1 8 1 1 20 15 7 11 2 974 1 4 5 4 0 7 14]
[ 9 1 3 24 11 0 0 6 36 3 6 3 2 0 1011 1 4 1 9 1 12]
[ 1 0 1 1 1 1 4 0 1 0 0 6 6 2 1 1031 8 8 0 7 16]
[ 3 4 1 0 4 1 0 0 5 0 0 3 6 0 5 18 1063 1 1 7 21]
[ 1 0 1 8 0 1 4 0 6 0 0 12 19 2 4 18 2 924 0 2 7]
[ 0 5 7 13 2 3 0 33 4 0 0 2 0 0 5 1 1 0 1015 2 8]
[ 0 2 0 1 1 1 9 6 0 0 0 12 6 1 2 10 8 0 0 1097 14]
[ 116 192 142 79 99 141 53 95 95 58 74 89 254 218 108 65 87 40 150 151 5485]]
Log file for this run: /home/madlabs/Documents/ai8x-training/logs/2023.01.28-232929/2023.01.28-232929.log
2428928000it [1:22:13, 492318.03it/s]5.屏幕适配
5.1 屏幕与控制板管脚对应关系
增加复位与背光控制管教宏定义
#define MXC_GPIO_PORT_RST MXC_GPIO2
#define MXC_GPIO_PIN_RST MXC_GPIO_PIN_6
#define MXC_GPIO_PORT_BL MXC_GPIO2
#define MXC_GPIO_PIN_BL MXC_GPIO_PIN_7
#define TFT_SPI MXC_SPI0 // SPI port to use for TFT display
#define TFT_SPI_PORT MXC_GPIO0 /**< GPIO port for SPI peripheral pins. */
#define TFT_SPI_PINS \
MXC_GPIO_PIN_5 | MXC_GPIO_PIN_6 | MXC_GPIO_PIN_7 /**< GPIO pins for SPI peripheral. */
#define TFT_DC_PORT MXC_GPIO2 /**< GPIO port for Data/Command signal. */
#define TFT_DC_PIN MXC_GPIO_PIN_7 /**< GPIO pin for Data/Command signal. */
#define TFT_SS_PORT MXC_GPIO2 /**< GPIO port for select signal. */
#define TFT_SS_PIN MXC_GPIO_PIN_4 /**< GPIO pin for select signal. */修改屏幕尺寸
#define screenWidth 240
#define screenHeight 240修改屏幕初始化接口
/* Initialize TFT display */
mxc_gpio_cfg_t gpio_in;
mxc_gpio_cfg_t gpio_bl;
mxc_gpio_cfg_t* gpio_in_p;
mxc_gpio_cfg_t* gpio_bl_p;
gpio_in_p = &gpio_in;
gpio_bl_p = &gpio_bl;
gpio_in.port = MXC_GPIO_PORT_RST;
gpio_in.mask = MXC_GPIO_PIN_RST;
gpio_in.pad = MXC_GPIO_PAD_PULL_UP;
gpio_in.func = MXC_GPIO_FUNC_IN;
/* Initialize TFT display */
MXC_TFT_Init(gpio_in_p, gpio_bl_p);调整游戏界面尺寸及控件位置
TFT_Print(buff, 50, 80, font_2, snprintf(buff, sizeof(buff), "Funpack 2-3"));
TFT_Print(buff, 50, 100, font_2, snprintf(buff, sizeof(buff), "SNAKE GAME"));
TFT_Print(buff, 10, 210, font_2, snprintf(buff, sizeof(buff), "PRESS PB1(SW1) TO GO"));
6.实验结果
7.遇到的问题
7.1开发环境用起来不方便
工程中用到的一些外设模块文件都是以库的形式添加到工程中，如果要移植其他的第三方库，要自己编写makefile。例如本项目中原示例工程所用TFT驱动型号为ili9341,如果改为其他型号的屏幕，就要通过修改makefile的形式，将源文件添加到工程内。
7.2工程配置中设置头文件路径不起作用
项目起初打算GUI采用lvgl，但是需要自己编写makefile，由于之前没有接触过，所以放弃，尝试将lvgl直接放到工程目录内，但是执行编译时，并未编译lvgl相关源文件，且包含lvgl相关头文件，也提示找不到头文件。
8.心得体会
非常高兴能够参与这次活动，由于之前没有接触过linux与机器学习相关的知识，通过这次活动，从安装系统，到安装各种开发环境依赖包，再到最后的模型训练，不仅掌握了linux系统的基本使用方法，对机器学习相关知识也有了一定认识。以前认为机器学习都需要在电脑上或者性能非常高的硬件上来实现，没想到在MCU上也可以实现。
附件下载
training-log-2023.01.28-232929.log
训练记录
funpack_snake_game.rar
工程文件
团队介绍
王者
评论
0 / 100
查看更多
猜你喜欢
Shark：基于树莓派PICO移植游戏-贪吃蛇使用MicroPython进行移植贪吃蛇游戏，游戏控制为MMA7660三轴姿态传感器，光电旋转编码器旋转控制贪吃蛇前进速度
shark_xin1430
基于micro python的复古游戏贪吃蛇改写与移植本项目基于ZP2040开发板引脚及相关特性，通过修改相关参数与引脚，对控制方法进行相应修改与优化，将适用于其他开发板的贪吃蛇游戏移植到ZP2040上并成功运行。
ax571375
【MAX78000第二季】MAX78000FTHR-快速实现超低功耗、人工智能 (AI) 方案MAX78000FTHR 是一个快速开发平台，可帮助工程师使用 MAX78000 Arm®Cortex®快速实现超低功耗、人工智能 (AI) 解决方案，具有集成卷积神经网络加速器的 M4F 处理器。
Lucia6864
