
Identify hardware requirements for deep learning on a Raspberry Pi, including model B Plus, a two-amp power supply, a device to accelerate object detection, and a 32 GB SD card.
Follow essential course navigation tips, access supplemental code in the GitHub repo and FAQs, and seek help via the CUNY forums, Stack Overflow, Q&A forums, and Udemy support.
Download Raspbian lite, then write with Win32DiskImager to a micro SD card. Meanwhile download SSH terminal program - putty.
Setup Wi-Fi access without a monitor and keyboard/mouse.
Find Raspberry Pi on your local network.
Login with putty.
Initial setup:
- sudo raspi-config -> expand filesystem;
- sudo apt-get update
- sudo apt-get upgrade
+ https://arxiv.org/pdf/1611.10012.pdf
A short description about the latest object detections models, especially focus on YOLO, that is capable of running on embedded devices.
The installation process might require 1-3 hours.
We will upload the installed and prepared image to a public storage soon (check github), so if you want then you can skip this lecture.
You can change the MAKE_NJOBS to 2 in the ncsdk.conf, with this the installation process might faster (this will use 2 core instead of 1, using 4 core is not required)
MAKE_NJOBS=2
I attached the command list, you can use like me in the video.
Currently the darkflow is not compatible with some tensorflow version, that is why I recommend to use a specific version (v1.4)
The openCV installation is not necessary, but later we will add a content how to inference (=run) an Object detection model with openCV on the Raspberry Pi CPU (instead of Movidius)
Install and set up darkflow, a flow-based open source YOLO implementation, on a Raspberry Pi, clone the repository, and adjust the flow utils for conversion.
mymovidius_yolo_test link:
https://drive.google.com/file/d/13KK_coUsmk0OSyHD6JYRJV7nd-i98w5c
Explore how to detect gender and age from camera data using deep learning on Raspberry Pi, with code examples and notes on downloading and compiling components.
Explore recurrent neural networks theory, memory, and backpropagation through time, addressing vanishing and exploding gradients with gradient clipping and long term short term memory architectures.
Discover how to accelerate deep learning on Raspberry Pi and edge devices for augmented reality. Explore computer vision concepts, OpenCV, Unity integration, and edge AI to power immersive AR experiences.
Learn how we implemented Deep Learning Object Detection Models on Raspberry Pi and accelerated them with Intel Movidius Neural Compute Stick.
When we first got started in Deep Learning particularly in Computer Vision, we were really excited at the possibilities of this technology to help people. The only problem is, that image classification and object detection runs just fine on our expensive, power consuming and bulky Deep Learning machines. However, not everyone can afford or implement AI for their practical applications.
This is when we went searching for an affordable, compact, less power hungry alternative. Generally if we'd want to shrink our IoT and automation projects, we'd often look to the Raspberry Pi which is versatile computing solution for numerous problems. This made us ponder about how we can port out deep learning models to this compact computing unit. Not only that, but how could we run it at close to real-time?
Amongst the possible solutions we arrived at using the raspberry pi in conjunction with an AI Accelerator USB stick that was made by Intel to boost our object detection frame-rate. However it was not so simple to get it up and running. Implementing the documentation, we landed up with a series of bugs after bugs, which became a bit tedious.
After endless posts on forums, tutorials and blogs, we have documented a seamless guide in the form of this course; which will show you, step-by-step, on how to implement your own Deep Learning Object Detection models on video and webcam without all the wasteful debugging. So essentially, we've structured this training to reduce debugging, speed up your time to market and get you results sooner.
In this course, here's some of the things that you will learn:
Getting Started with Raspberry Pi even if you are a beginner,
Deep Learning Basics,
Object Detection Models - Pros and Cons of each CNN,
Setup and Install Movidius Neural Compute Stick (NCS) SDK,
Currently, the OpenVINO is available for Raspbian, so the NCS2 is already compatible with the Raspberry Pi, but this course is mainly for the Movidius (NCS version 1).
Run Yolo and Mobilenet SSD object detection models in recorded or live video
You also get helpful bonuses:
*OpenCV CPU inference
*Introduction to Custom Model Training
Personal help within the course
I donate my time to regularly hold office hours with students. During the office hours you can ask me any business question you want, and I will do my best to help you. The office hours are free. I don't try to sell anything.
Students can start discussions and message me with private questions. I answer 99% of questions within 24 hours. I love helping students who take my courses and I look forward to helping you.
I regularly update this course to reflect the current marketing landscape.
Get a Career Boost with a Certificate of Completion
Upon completing 100% of this course, you will be emailed a certificate of completion. You can show it as proof of your expertise and that you have completed a certain number of hours of instruction.
If you want to get a marketing job or freelancing clients, a certificate from this course can help you appear as a stronger candidate for Artificial Intelligence jobs.
Money-Back Guarantee
The course comes with an unconditional, Udemy-backed, 30-day money-back guarantee. This is not just a guarantee, it's my personal promise to you that I will go out of my way to help you succeed just like I've done for thousands of my other students.
Let me help you get fast results. Enroll now, by clicking the button and let us show you how to develop Accelerated AI on Raspberry Pi.