Asus Tinker Edge T - Object Detection Tutorial (opencv, gstreamer)

Tutorial: How to perform Object Detection Inference on the Asus Tinker Edge T (ATET).

Download the Google Coral Github which contains example Inference code for gstreamer, opencv, pygame, and raspicam. This video covers opencv and gstreamer.

Upon booting into Mendel Linux, open a weston-terminal and install Git:
sudo apt-get install git

Then clone the Google Coral Github:
sudo git clone https://github.com/google-coral/examp...

Change Directory to your local repo:
cd examples_camera

Download the Tensorflow Lite / tflite Edge TPU Object Detection models:
sudo sh download_models.sh
e.g., mobilenet v1, mobilenet v2, ssd v1, ssd v2, inception v1, inception v3, inception v4, coco, edgetpu

Install python3-pycoral:
sudo apt-get install python3-pycoral

Change Directory into OpenCV, Install Requirements, and run detect.py:
cd opencv
sudo sh install_requirements.sh
python3 detect.py

For gstreamer:
cd gstreamer (from ./examples_camera)
sudo sh install_requirements.sh
python3 detect.py

If you get this VIDIOC_QUERYCTRL: No such device error - it means the detect.py scripts are looking for your webcam on the wrong USB port. To find which USB port your webcam is connected to, type in:
v4l2-ctl --list-devices

You should see your webcam model and the port number (e.g., C922 Pro on dev/video2)

Then go back to your Python script (detect.py) and update the camera_idx variable:
sudo nano python3
(search for the camera_idx, default = 0) and change the number to whatever port your webcam is plugged into - mine was video2 so I'll change it to default = 2.

If you can't exit gstreamer detect.py, open TTY: CTRL ALT F2
Type in your username and password (default is mendel, mendel)
Search for all processes & find the Process ID that's running detect.py: ps -aux | more
Once you've found it, kill it with: kill PROCESS_ID
Exit TTY: CTRL ALT F7

Asus Tinker Edge T spec sheet:
CPU: Asus Tinker Edge T SoC 1.5GHz Quad Core Cpu (NXP i.MX 8M)
RAM: 1GB LPDDR4
GPU: GC7000 Lite
TPU: Google Edge TPU ML accelerator
Display: HDMI & 22-pin MIPI DSI
Storage: 8GB eMMC, MicroSD slot
Connectivity: RTL8211F-CG Gb LAN, 802.11 a/b/g/n/ac wireless & Bluetooth 4.2
USB: 2x USB 3.2 Gen1 Type-A, USB 3.2 Gen1 Type-C OTG
Camera Interface: 2x 24-pin MIPI CSI-2
1 x 40-pin headers:
up to 28 x GPIO pins
up to 1 x SPI bus
up to 2 x I2C bus
up to 2 x UART
up to 3 x PWM
up to 1 x PCM/I2S
2 x 5V power pins
2 x 3.3V power pins
8 x ground pins
1 x Boot mode switch
1 x 2-pin Reset header
1 x 2-pin DC Fan header
Power: 12~19v DC Power Input Jack (5.5/2.5mm)

The Asus Tinker Edge T is part of the Tinker Board Single Board Computer (SBC) family. However, this board that can perform Computer Vision Artificial Intelligence (AI) and Deep Learning inference (i.e., Image Classification, Object Detection, Segmentation, Pose Estimation) with the same level of performance that you'd find in an expensive desktop computer. It can achieve this because it contains a special chip called an "Edge TPU" (Tensor Processing Unit) that specialises in calculating AI and Deep Learning algorithms which can also be found in a Google Coral Dev Board.
----------
Subscribe to my YouTube channel:
https://www.youtube.com/edgecate?sub_...
----------
#Asus, #TinkerBoard, #TPU