AI邊緣運算實作: TensorFlow Lite for MCU
https://bit.ly/3j2fIIt
[1]python程式設計
https://bit.ly/359cz4m
[2]AI機器學習&深度學習
http://bit.ly/2KDZZz4
[3]TensorFlow Lite for MCU
https://bit.ly/3j2fIIt
Tiny ML for spark Fun Edge
https://www.ittraining.com.tw/ittraining/it-elearning/el-ai/ai-tensorflow-lite-for-mcu
TensorFlow Lite for MCU正是專為邊緣裝置設計的TensorFlow模型預測框架,是TensorFlow的精簡版本,讓開發者可以在物聯網與嵌入式裝置中部署微型機器學習模型。 本課程將教授AI模型如何佈署於微控制器,包含模型訓練、模型最佳化以及TensorFlow Lite框架的程式開發等。此外,在實作上以Sparkfun edge board (ARM cortex M4)為例,說明如何以TensorFlow Lite 進行微控制器上面的人工智慧開發專案,包含人臉偵測、關鍵字的字詞偵測、姿態識別、異常偵測等。
The OpenVINO toolkit enables deep learning inference on embedded devices by supporting heterogeneous execution across Intel CPUs, GPUs, FPGAs, and Movidius VPUs. It includes optimized computer vision functions and pre-trained models. The toolkit provides a model optimizer to optimize models for size and speed and an inference engine to run models across different hardware accelerators. Developers can use OpenVINO to easily deploy CNN-based applications with real-time performance on embedded systems.
This document discusses stepper motors, including their types, resolutions, gear ratios, coils, phases, and drive methods. It provides details on a specific 28BYJ-48 5V unipolar stepper motor, including its specifications. It explains half-step sequencing and how stepper motor movement is controlled by pulse signals. Code examples are given for controlling a stepper motor with a Raspberry Pi. The document contrasts unipolar and bipolar stepper motors and their drive methods.
This document discusses the basics of artificial neural networks including multi-layer perceptrons (MLPs). It explains that MLPs use multiple hidden layers between the input and output layers to extract meaningful features from the data. The document also covers topics like training neural networks using backpropagation and stochastic gradient descent, the use of mini-batches to speed up training, and common activation and loss functions.
https://youtu.be/RHvROP94qZ0
AI邊緣運算實作: TensorFlow Lite for MCU
https://bit.ly/3j2fIIt
[1]python程式設計
https://bit.ly/359cz4m
[2]AI機器學習&深度學習
http://bit.ly/2KDZZz4
[3]TensorFlow Lite for MCU
https://bit.ly/3j2fIIt
Reinforcement learning allows an agent to learn how to behave through trial-and-error interactions with an environment. The agent takes actions in a state and receives rewards, learning through experience which actions maximize total rewards. The agent learns a policy using a Q-table that represents the estimated utility of taking an action in a given state. Initially the agent explores randomly, but over time exploits what it has learned from the Q-table to select the highest-valued actions. The Q-learning algorithm iteratively updates the Q-table values using the Bellman equation to improve its estimates of the best actions.
The document discusses Linux device trees and how they are used to describe hardware configurations. Some key points:
- A device tree is a data structure that describes hardware connections and configurations. It allows the same kernel to support different hardware.
- Device trees contain nodes that represent devices, with properties like compatible strings to identify drivers. They describe things like memory maps, interrupts, and bus attachments.
- The kernel uses the device tree passed by the bootloader to identify and initialize hardware. Drivers match based on compatible properties.
- Device tree files with .dts extension can be compiled to binary blobs (.dtb) and overlays (.dtbo) used at boot time to describe hardware.
9. 2-2. Fusion Tables (2)
Google的官方定義 :
Google Fusion Tables is a web application used for sharing, visualizing, and
publishing tabular data.
You can upload your own CSV, KML, ODS, XLS, or Google Spreadsheet data to a
Fusion Tables table.
Once your data is in Fusion Tables, you can collaborate on it with others in real
time, publish it for Google Search, create map and chart visualizations for
private use or for embedding on websites, filter it according to specific criteria,
and update the data behind your visualizations or filters at any time.
10. 2-2. Fusion Tables (3)
Google Cloud
Google
各種雲端應用服務Google
Developers Console
(提供各種API)
Fusion Table
API
Fusion Table
user
Browser裝置應用程式
Google針對各種雲端應用服務
提供相對應的API
針對Fusion Tables, 其最新版
API為Fusion Tables API v2.0
11. 2-2. Fusion Tables (4)
Google針對各種雲端應用服務所提供的API, 可在Google API Explorer中找到:
https://developers.google.com/apis-explorer/#p/
12. 2-2. Fusion Tables (5)
Google對於 Fusion Tables API 的官方定義 :
The Fusion Tables API allows you to use HTTP requests to programmatically
to perform these tasks, which are also available in the Fusion Tables web
application:
1. create and delete tables
2. read and modify table metadata such as table and column names and
column types
3. insert, update, and delete the rows in a table
4. create, update, and delete settings for certain visualizations
5. query the rows in a table
13. 2-2. Fusion Tables (6)
Google API 所共用的
認證機制為
“OAuth 2.0”
非Andriod或iOS的行動裝
置, 可採用
“OAuth 2.0 for Devices”
[ 圖片來源: Google 網站
https://developers.google.com/identity/protocols/OAuth2ForDevices ]