posted Nov 27, 2020, 11:32 PM by MUHAMMAD MUN`IM AHMAD ZABIDI
- Meyer, Matthias, Lukas Cavigelli, and Lothar Thiele. "Efficient convolutional neural network for audio event detection." arXiv preprint arXiv:1709.09888 (2017). \url{https://arxiv.org/pdf/1709.09888.pdf}
- Abdrakhmanov, Vali Kh, Renat B. Salikhov, and Konstantin V. Vazhdacv. "Development of a sound recognition system using STM32 microcontrollers for monitoring the state of biological objects." 2018 XIV International Scientific-Technical Conference on Actual Problems of Electronics Instrument Engineering (APEIE). IEEE, 2018. \url{https://ieeexplore.ieee.org/document/8545278} <-- privileged
- Crocioni, Giulia, et al. "Li-Ion Batteries Parameter Estimation With Tiny Neural Networks Embedded on Intelligent IoT Microcontrollers." IEEE Access 8 (2020): 122135-122146.\url{https://ieeexplore.ieee.org/document/9133084}
- Li, Hao, et al. "Design of micro-automatic weather station for modern power grid based on STM32." The Journal of Engineering 2017.13 (2017): 1629-1634.\url{https://ieeexplore.ieee.org/iel7/7864294/8311003/08311183.pdf}
- Luna-Perejón, Francisco, et al. "Low-Power Embedded System for Gait Classification Using Neural Networks." Journal of Low Power Electronics and Applications 10.2 (2020): 14. \url{https://www.mdpi.com/2079-9268/10/2/14/pdf}
- Sakr, Fouad, et al. "Machine Learning on Mainstream Microcontrollers." Sensors 20.9 (2020): 2638. \url{https://www.mdpi.com/1424-8220/20/9/2638/pdf}
- Chaber, Patryk, and Maciej Ławryńczuk. "Fast analytical model predictive controllers and their implementation for STM32 ARM microcontroller." IEEE Transactions on Industrial Informatics 15.8 (2019): 4580-4590. \url{https://www.researchgate.net/profile/Maciej_awrynczuk/publication/330372035_Fast_Analytical_Model_Predictive_Controllers_and_Their_Implementation_for_STM32_ARM_Microcontroller/links/5daf1a21299bf111d4bfbd15/Fast-Analytical-Model-Predictive-Controllers-and-Their-Implementation-for-STM32-ARM-Microcontroller.pdf}
- Lin, Yi-Bing, et al. "Morsocket: an expandable iot-based smart socket system." IEEE Access 6 (2018): 53123-53132. \url{https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=8466570}
- Ji, Li, Shao Qiongling, and Wang Shengjun. "Design of Embedded Network Voice Communication Terminal Based on STM32 and μCOSIII." MATEC Web of Conferences. Vol. 173. EDP Sciences, 2018. \url{https://www.matec-conferences.org/articles/matecconf/pdf/2018/32/matecconf_smima2018_03027.pdf}
- Canta, Riccardo. "Design of an optimized audio framework for portable digital MEMS microphones evaluation." (2012). Thesis. Politecnico di Milano. \url{https://www.politesi.polimi.it/bitstream/10589/72062/1/Canta_Riccardo_754573.pdf}
|
posted Nov 15, 2020, 2:09 AM by MUHAMMAD MUN`IM AHMAD ZABIDI
Fig. 2 from R. Boutaba et al., “A comprehensive survey on machine
learning for networking: evolution, applications and research
opportunities,” J. Internet Serv. Appl., vol. 9, no. 1, p. 99, 2018.
Problem categories that benefit from machine learning.
a Clustering.
b Classification.
c Regression.
d Rule extraction
|
posted Nov 15, 2020, 2:06 AM by MUHAMMAD MUN`IM AHMAD ZABIDI
[
updated Nov 15, 2020, 2:06 AM
]
Fig. 3 from R. Boutaba et al., “A comprehensive survey on machine learning for networking: evolution, applications and research opportunities,” J. Internet Serv. Appl., vol. 9, no. 1, p. 99, 2018.
|
posted Nov 15, 2020, 1:59 AM by MUHAMMAD MUN`IM AHMAD ZABIDI
[
updated Nov 15, 2020, 2:03 AM
]
Fig. 5 from R. Boutaba et al., “A comprehensive survey on machine learning for networking: evolution, applications and research opportunities,” J. Internet Serv. Appl., vol. 9, no. 1, p. 99, 2018.
|
posted Nov 15, 2020, 1:56 AM by MUHAMMAD MUN`IM AHMAD ZABIDI
[
updated Nov 15, 2020, 1:57 AM
]
Table 3 from M. G. S. Murshed, C. Murphy, D. Hou, N. Khan, G. Ananthanarayanan, and F. Hussain, “Machine Learning at the Network Edge: A Survey,” arXiv Prepr. arXiv1908.00080, pp. 1–28, 2019.
|
Device
|
GPU
|
CPU
|
RAM
|
Flash
memory
|
Power
consumption
|
Example
applications
|
|
Raspberry Pi
|
400MHz
VideoCore IV
|
Quad
Cortex A53 @ 1.2GHz
|
1 GB SDRAM
|
32 GB
|
2.5 Amp
|
video analysis [78, 114]
|
|
Coral Dev Board (Edge TPU)
|
GC7000 Lite
Graphics + Edge TPU coprocessor
|
Quad Cortex-A53, Cortex-M4F
|
1 GB LPDDR4
|
8 GB LPDDR4
|
5V DC
|
image processing [18]
|
|
SparkFun Edge
|
-
|
32-bit ARM
Cortex-M4F 48MHz
(with 96MHz
burst mode) processor
|
384KB
|
1MB
|
6uA/MHz
|
speech recognition [31]
|
|
Jetson TX1
|
Nvidia
Maxwell 256 CUDA
cores
|
Quad ARM A57/2 MB L2
|
4 GB
64 bit LPDDR4
25.6 GB/s
|
16 GB eMMC, SDIO, SATA
|
10-W
|
video, image analysis [68] [61]
robotics [33]
|
|
Jetson TX2
|
Nvidia Pascal
256 CUDA
cores
|
HMP Dual
Denver 2/2 MB L2 +
Quad ARM
A57/2 MB L2
|
8 GB
128 bit LPDDR4
59.7 GB/s
|
32 GB eMMC, SDIO, SATA
|
7.5-W
|
video, image analysis [68], [92]
robotics [24]
|
|
Intel Movidius Neural Compute Stick
|
High Performance VPU
|
Myriad 2 Vision Processing Unit
|
1 GB
|
4 GB
|
2 trillion
16-bit operations per second within
500 mW
|
classification [73] computer
vision [14, 43]
|
|
ARM ML
|
-
|
ARM ML
processor
|
1 GB
|
-
|
4 TOPs/W
(Tera Operations)
|
image, voice recognition [107]
|
|
RISC-V GAP8
|
-
|
nona-core
32-bit RISC-V
microprocessor
@250 MHz
|
16 MiB SDRAM
|
-
|
1 GOPs/mW
|
image, audio processing [35]
|
|
OpenMV
Cam
|
-
|
ARM 32-bit
Cortex-M7
|
512KB
|
2 MB
|
200mA
@ 3.3V
|
image
processing [6]
|
|
BeagleBone AI
|
-
|
Cortex-A15
Sitara AM5729 SoC with 4 EVEs
|
1 GB
|
16 GB
|
-
|
computer vision [25]
|
|
EMC3531
|
-
|
ARM Cortex-M3
NXP Coolflux DSP
|
-
|
-
|
-
|
audio, video analysis38
|
|
posted Nov 15, 2020, 1:53 AM by MUHAMMAD MUN`IM AHMAD ZABIDI
[
updated Nov 15, 2020, 1:55 AM
]
Table 3 from M. G. S. Murshed, C. Murphy, D. Hou, N. Khan, G. Ananthanarayanan, and F. Hussain, “Machine Learning at the Network Edge: A Survey,” arXiv Prepr. arXiv1908.00080, pp. 1–28, 2019.
Framework | Core language | Interface | Part running on the edge | Example applications | TensorFlow Lite (Google) | C++ Java | C/C++ Java | TensorFlow Lite NN API | computer vision [109], speech recognition [42, 1] | Caffe2 Caffe2Go (Facebook) | C++ | Android iOs | NNPack | image analysis, video analysis [53] | Apache MXNet | C++ Python R | Linux MacOS Windows | Full Model | object detection, recognition [78] | Core ML2 (Apple) | Python | iOS | CoreML | image analysis [16] NLP [105] | ML Kit (Google) | C++ Java | Android iOs | Full Model | image recognition, text recognition, bar-code scaning [26] | AI2GO | C, Python Java, Swift | Linux macOs | Full Model | object detection, classification [5] | DeepThings | C/C++ | Linux | Full Model | object detection [119] | DeepIoT | Python | Ubilinux | Full Model | human activity recognition, user identification [116] | DeepCham | C++ Java | Linux Android | Full Model | object recognition [62] | SparseSep | - | Linux Android | Full Model | mobile object recognition, audio classification [15] | Edgent | - | Ubuntu | Major part of the DNN | image recognition [63] |
17 |
posted Nov 11, 2020, 7:02 AM by MUHAMMAD MUN`IM AHMAD ZABIDI
[
updated Nov 11, 2020, 7:15 AM
]
Some embedded/IoT projects that require extensive processing and low power at the same time. Like those that would take advantage of STM32F4/F7 DSP, FPU and SIMD, plus CMSIS-NN and CMSIS-DSP libraries. The selected are those that can be downloaded directly. I skip those that require authentication.
- Amoh, J. and Odame, K. (2016). Deep neural networks for identifying cough sounds. IEEE transactions on biomedical circuits and systems, 10(5):1003–1011.
Available https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=7570164.
- Amoh, J. and Odame, K. M. (2019). An optimized recurrent unit for ultra-low-power keyword spotting. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies,3(2):1–17.
Available https://dl.acm.org/doi/pdf/10.1145/3328907.
- Lai, L., Suda, N., and Chandra, V. (2018). CMSIS-NN:Efficient neural network kernels for ARM Cortex-M CPUs. arXiv preprint arXiv:1801.06601.
Available https://arxiv.org/pdf/1801.06601.pdf.
- Liu, P.-X., Chen, Y.-j., Jiang, B.-H., and Zhang, X. (2016). Design of the data acquisition system based on STM32. In 2016 Inter-national Conference on Computational Science and Engineering (ICCSE2016). Atlantis Press.
Available https://download.atlantis-press.com/article/25862099.pdf.
- Xie, Y., Su, X., He, Y., Chen, X., Cai, G., Xu, B., and Ye,W. (2017). STM32-based vehicle data acquisition system for internet-of-vehicles. In 2017 IEEE/ACIS 16th International Conference on Computerand Information Science (ICIS), pages 895–898. IEEE.
Available https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=7960119.
- Zhang, Y. and Li, L. (2014). Smart home system basedon Zigbee network and STM32f407 microprocessor. The Open Cybernetics& Systemics Journal, 8(1).
Available https://benthamopen.com/contents/pdf/TOCSJ/TOCSJ-8-651.pdf.
- Zhang, Y., Suda, N., Lai, L., and Chandra, V. (2017). Hello Edge: Keyword spotting on microcontrollers. arXiv preprintarXiv:1711.07128.
Available https://arxiv.org/pdf/1711.07128.pdf.
- Fezari, Mohamed, Mounir Bousbia-Salah, and Mouldi Bedda. Microcontroller Based Heart Rate Monitor. International Arab Journal of Information Technology (IAJIT) 5.4 (2008).
Available http://www.academia.edu/download/49169562/11-198.pdf.
|
posted Oct 5, 2020, 8:15 AM by MUHAMMAD MUN`IM AHMAD ZABIDI
[
updated Oct 17, 2020, 9:52 PM
]
- A Comprehensive Guide to Machine Learning,
By Soroush Nasiriany, Garrett Thomas, William Wang, Alex Yang.
Department of Electrical Engineering and Computer Sciences, University of California, Berkeley. Dated June 24, 2019.
https://t.co/iZsREscT9v - The Math of Machine Learning - Berkeley University Textbook
By Garret Thomas.
needed for an introductory class in machine learning, which at UC Berkeley is known as CS 189/289A.
https://www.datasciencecentral.com/profiles/blogs/tutorial-the-math-of-machine-learning-berkeley-university - Foundations of Data Science
By Avrim Blum, John Hopcroft, and Ravindran Kannan (2018).
from Microsoft Research Lab
https://www.cs.cornell.edu/jeh/book.pdf - Machine Learning from Scratch
By Danny Friedman
author is a stats major at Harvard
https://dafriedman97.github.io/mlbook/content/introduction.html
More books here
|
posted Sep 27, 2020, 4:03 PM by MUHAMMAD MUN`IM AHMAD ZABIDI
http://viz.smultron.org/r/spectrograms/#more-94
|
posted Sep 22, 2020, 4:43 PM by MUHAMMAD MUN`IM AHMAD ZABIDI
[
updated Sep 22, 2020, 5:12 PM
]
Original paper:
Results using CIFAR-10 below.
CIFAR-10 CNN as implemented by CMSIS-NN.
Performance of CMSIS-NN over baseline CNN written using arm_conv in CMSIS-DSP.
Platform: NUCLEO-F746ZG mbed board with an Arm Cortex-M7 core running at 216 MHz. Note: using state-of-art DNN, researchers has achieved ≥93% of accuracy on CIFAR-10.
Performance on STM32 compared with other SoCs, as reported by AiOTA Labs.
The players: - STM32F7 from STmicro. Up to 216 MHz.
- GAP8 from Greenwaves with CNN benchmarks. Max freq 175 MHz "Cluster", 250 MHz "Fabric Controller". RISC-V core.
- i.Mx 6ULL from NXP. This guy goes up to 900 MHz.
Looks like the GAP8 is many times more power-efficient than Cortex. At 10 FPS, the GAP8 needs 3.7 mW versus 60 mW on STM32. But it comes at a price. Very heavy price. The cost of an Arduino-compatible GAP8 board is 100,00€. And it's not running CMSIS-NN either. CMSIS-NN is for ARM processors only.
Back to CMSIS-NN.
|
|
