Train an ESP32-CAM for Object Detection using a free Edge Impulse account. All without writing any code!
Article with full instructions: dronebotworkshop.com/esp32-ob...
More articles and tutorials: dronebotworkshop.com
Join the conversation on the forum: forum.dronebotworkshop.com
Subscribe to the newsletter and stay in touch: dronebotworkshop.com/subscribe/
Object Detection is a Machine Learning (ML) technique that uses a computer with a camera to identify classes of objects it has been trained to recognize. Although this used to require massive computers and large datasets, it is now possible to perform decent object detection using microcontrollers.
While most microcontroller-based object detection projects use 64-bit chips like the powerful Kendryte K210, you can also use a 32-bit ESP32 to get the job done. And that’s exactly what we will do today - take a 9-dollar ESP32-CAM board and train it to detect objects.
I’ll be using a lantern battery and Robie, a 40-year-old Radio Shack robot, as my two “test objects” We’ll train a model using the powerful online neural network services of Edge Impulse; you’ll need an account, but don’t worry, it’s completely free.
Once we have our model trained, we’ll export it in the form of an Arduino Library. The library even includes a sample sketch we can use to test our model; no code to write! I’ll try it on an ESP32-CAM and ESP-EYE board.
Here is the Table of Contents for today's video:
00:00 - Introduction
02:34 - Object Detection
06:27 - Edge Impulse
08:15 - Workflow
10:47 - Look at ESP32-CAM & ESP-EYE
12:48 - Capturing Images - Webcam Setup
14:34 - Edge Impulse Setup
15:39 - Image Capture
17:57 - Label Images
20:32 - Create an Impulse
24:52 - Export to an Arduino Library
25:25 - Importing Library to Arduino IDE
27:41 - Testing with ESP-EYE
29:43 - Testing with ESP32-CAM
32:04 - EloquentESP32CAM Library
34:14 - Collect Images Sketch
35:35 - Capturing Images - ESP32-CAM
39:12 - Import & Label with Edge Impulse
42:46 - Build & Deploy Impulse
46:45 - Testing with ESP32-CAM
48:28 - EloquentArduino Code & Demo
51:32 - Conclusion
53:25 - Robie the Robot!
I will actually show you two methods of capturing images of your subject(s) - using a webcam or using the ESP32-CAM board itself. I’ll also introduce you to a library that can make doing all of this a lot easier.
The results are good, good enough to consider using the ESP32-CAM as an inexpensive object detection sensor.
Hope you enjoy the video!