Share E-Book
Scan to open this page

Scan with your phone to open this page

AuthorDaniel Situnayake

Deep learning networks are getting smaller. Much smaller. The Google Assistant team can detect words with a model just 14 kilobytes in size—small enough to run on a microcontroller. With this practical book you’ll enter the field of TinyML, where deep learning and embedded systems combine to make astounding things possible with tiny devices. Pete Warden and Daniel Situnayake explain how you can train models small enough to fit into any environment. Ideal for software and hardware developers who want to build embedded systems using machine learning, this guide walks you through creating a series of TinyML projects, step-by-step. No machine learning or microcontroller experience is necessary. * Build a speech recognizer, a camera that detects people, and a magic wand that responds to gestures * Work with Arduino and ultra-low-power microcontrollers * Learn the essentials of ML and how to train your own models * Train models to understand audio, image, and accelerometer data * Explore TensorFlow Lite for Microcontrollers, Google’s toolkit for TinyML * Debug applications and provide safeguards for privacy and security * Optimize latency, energy usage, and model and binary size

AI Reading Assistant

Whole-book reading guide from stratified index samples; jump to passages in the text

Brief outline
【One-Line Pitch】 Deep learning networks are getting smaller. Much smaller. The Google Assistant team can detect words with a model jus… 【Book Arc】 - **Opening (~0%–12%)**: The Google Assistant team can detect words with a model just 14 kilobytes in size—small enough to run on a microcontroller.; obile phone’s main CPU and so draws hundreds of milliwatts. - **Early (~12%–35%)**: If our model isn’t working in production, we’ll need to do some troubleshooting.; This can be a useful metric when discussing model size and complexity. - **Middle (~35%–65%)**: int brightness (int)(127.5f * (y_value 1));; 0x1 0x49f40003 0xffffffff [...lots more 0xffffffff...] python -c 'import fileinput, glob; - **Late (~65%–88%)**: This is, essentially, the smart vision sensor we described a little earlier.; red as a flat in-memory array, it doesn’t really make much difference whether it is wrapped in unnecessary extra dimensions. - **Ending (~88%–100%)**: l show you how to use it with the model we’ve trained.; s, where the filter window doesn’t cover an actual value, are padded with zeros. 【Key Takeaways】 - **The Google Assistant t…** (Opening): The Google Assistant team can detect words with a model just 14 kilobytes in size—small enough to run on a microcontroller. - **obile phone’s main CPU…** (Opening): obile phone’s main CPU and so draws hundreds of milliwatts. - **Evaluate and troublesh…** (Opening): Evaluate and troubleshoot work effectively, these floating-point values need to be similar in size to one another. - **If our model isn’t wor…** (Early): If our model isn’t working in production, we’ll need to do some troubleshooting. - **This can be a useful m…** (Early): This can be a useful metric when discussing model size and complexity. - **converter to create an…** (Early): converter to create and save two new versions of our model. 【Reading Tips】 - Use Passage locations below to jump into the text and set reading anchors - If this is a brief outline, click Regenerate (top right) for a synthesized guide 【Coverage Limits】 Compressed outline without the model (~32 index chunks). Full structured guide needs AI available.
Page 11
obile phone’s main CPU and so draws hundreds of milliwatts. Keeping one running even for a few days requires a battery similar to a smartphone’s, making it d...
View in text
Excerpt 2
run in Colab, you’ll see that a 1 is now displayed in the upper-left corner when it is no longer selected, as illustrated in Figure 4-7. This number is a cou...
View in text
Excerpt 3
EXPECT_NEAR(0.141, value, 0.05); input->data.f[0] = 5.; interpreter.Invoke(); value = output->data.f[0]; TF_LITE_MICRO_EXPECT_NEAR(-0.959, value, 0.05); Note...
View in text
Excerpt 4
est.cc Shows how to run inference on spectrogram data and interpret the results audio_provider_test.cc Shows how to use the audio provider feature_provider_m...
View in text
Excerpt 5
smoothing, making the trends more obvious. Figure 8-12. TensorBoard’s Smoothing slider Figure 8-13 shows the same graphs with a higher level of smoothing. Th...
View in text
Excerpt 6
ery device-specific, and if you want to use it in your own code you can just copy and paste it. It configures the Arduino’s hardware to communicate with the...
View in text
Excerpt 7
l show you how to use it with the model we’ve trained. The majority of the code is preparing example images to feed into the trained network so that the rang...
View in text
Excerpt 8
g all of the terminal’s output to a file named output.txt: specialized version of TensorFlow Lite aimed just at these embedded platforms. The goal was to reu...
View in text
Tags
AI categories
AI
ISBN: 1492052043
Publisher: O'Reilly Media
Publish Year: 2019
Language: English
File Format: PDF
File Size: 11.8 MB
Text Preview (First 20 pages)
Registered users can read the full content for free

Register as a Gaohf Library member to read the complete e-book online for free and enjoy a better reading experience.

Generating text preview…