【PLC based on IoT】---Testing RPI-400 performance based on AI

cpsiot

【PLC based on IoT】---Testing RPI-400 performance based on AI [Copy link]

 

It is a bit of a waste to use RPI-400 as the HMI part. Today, I will test the AI performance. If it is OK, I can do the AI recognition part locally. Since it is IOT, how can I leave AI? !

Hardware used for testing:

Open (the following operations are based on the latest version of Raspberry Pi Bullseye 64-bit):

sudo apt install -y \
swig libjpeg-dev zlib1g-dev python3-dev \
unzip wget python3-pip curl git cmake make

sudo apt install python3-pyaudio

pip3 install -i https://pypi.tuna.tsinghua.edu.cn/simple/ numpy==1.22.3

curl -OL https://ghproxy.com/https://github.com/PINTO0309/TensorflowLite-bin/releases/download/v2.9.0/tflite_runtime-2.9.0-cp39-none-linux_aarch64.whl

sudo pip3 install --upgrade tflite_runtime-2.9.0-cp39-none-linux_aarch64.whl

After installation, open python3 in the terminal and enter. If there is no error, the installation is successful.
import tflite_runtime.interpreter as tflite

Download and run the object detection test program of TensorFlow Lite:

git clone https://ghproxy.com/https://github.com/tensorflow/examples --depth 1

cd examples/lite/examples/object_detection/raspberry_pi

sh setup.sh

Enter the following two addresses in the browser address bar to automatically download, and name the downloaded files as efficientdet_lite0.tfliteand efficientdet_lite0_edgetpu.tfliteput them in setup.shthe same directory as required .

https://tfhub.dev/tensorflow/lite-model/efficientdet/lite0/detection/metadata/1?lite-format=tflite

https://storage.googleapis.com/download.tensorflow.org/models/tflite/edgetpu/efficientdet_lite0_edgetpu_metadata.tflite

After installation, run the program:

python3 detect.py --model efficientdet_lite0.tflite

Use VNC to connect to the Raspberry Pi desktop:

The mouse and keyboard were successfully recognized. AI localization preparation is complete.

Jacktang

Want to know the specific principle of identifying the mouse and keyboard

cpsiot

Jacktang posted on 2022-7-24 10:08 I want to know the specific principle of identifying the mouse and keyboard

The trained model library efficientdet_lite0.tflite in DEMO contains keyboard and mouse, so it can be recognized. If it is not in the library, it cannot be recognized. You need to retrain it yourself. For details, please refer to the TensorFlow Lite official website.