Overview

Parameters

• Refresh time: The refresh time is based on experimental test data, and there may be errors in the actual refresh time. The actual results shall prevail. There will be a flickering effect during the global refresh process, which is a normal phenomenon.
  

Picture Production

• To make a scatter plot with PS and other tools, please refer to: E-Paper Floyd-Steinberg
• Use the tools we have written to convert images, applications and source codes: Six-color dithering image conversion tool
  • The program converts images to appropriate resolutions and optimizes them using dithering algorithms, supports most formats of images, recognizes the orientation of images and automatically stretches and crops them
  • How to use: Drag and drop the picture directly onto "convert.exe" to convert a single image, or put all the images into a folder and then double-click "converterTo6color_all.cmd" for batch conversion
  • Note: This program is developed and open source for the convenience of customers only, and we do not provide technical support for it

Tools for Windows

• Conversion of a single image
  

Drag the image onto convert.exe

• Batch conversion of images
  

Put the images, convert.exe, and converterTo6color_all.cmd in a folder, and then double-click converterTo6color_all.cmd

Tools for MAC

Put the images, convert, and converterTo6color_all in a folder
Open the terminal and enter the corresponding folder

cd xxxx/xxx/xxx/

• Conversion of a single image
  

Enter the command to grant the convert executable permission

sudo chmod +x convert

Image conversion

./convert image name

• Batch conversion of images
  

You need to perform permission operations on the convert file first
Enter a command to give converterTo6color_all.sh executable permissions

sudo chmod +x converterTo6color_all.sh

Image conversion

./converterTo6color_all.sh

Introduction to Pre-built Systems

• The large AI models are pre-set up and ready to use
• The default time zone 0 is used, which can be modified through the Time Zone Settings section
• The default update time is 8 o'clock sharp, which can be modified through the Scheduled Refresh section
• It is recommended to clear the corresponding logs regularly, refer to Script Usage and Scheduled Refresh
• Username: pi
• Password: raspberry
• Host name: zero2w

AI Model Construction

Refer to Dylan's open-source program on GitHub for building: https://github.com/dylski/PaperPiAI
The construction is relatively complex, and the process requires networking. You can also use the system we have already built

Hardware Preparation

• 32GB memory card, about 20GB of content will be added during the construction process
• Recommended system: Raspberry Pi OS (64-bit)
• Raspberry Pi Zero 2 W (PS: I used this directly for installation)
• ETH/USB HUB HAT (Optional, PS: Mainly used for accessing the network cable, downloading via WiFi is not as stable as downloading via the network cable)

Software Construction

• Flash the system to the memory card
• Update the source and update the software

sudo apt-get update
sudo apt-get upgrade

• Install git, download the repository

sudo apt install git

• Download reference repository (optional)

git clone https://github.com/dylski/PaperPiAI.git
cd PaperPiAI
scripts/install.sh  # You can install it directly using the installation script, or by executing the commands in the shell file step by step

• Install the necessary tools

sudo apt-get -y install tmux vim
sudo apt-get -y install cmake
sudo apt-get -y install python3-dev python3-venv python3-pip
sudo apt-get -y install imagemagick
sudo apt-get -y install git git-lfs
sudo apt-get -y install libopencv-dev  python3-opencv

• Create a folder and enter (I directly downloaded the repository here, so I'm using this directly)

cd PaperPiAI

• Create a virtual environment (only create it once; if created later, some software packages need to be re-downloaded)

python3 -m venv venv
PS: Why install a virtual environment?
    Under the official requirements of Python, some python libraries can only be installed in a virtual environment
    The latest Raspberry Pi system also did the same

• Enter the virtual environment (exit command: deactivate)

. venv/bin/activate

• Install some software

python -m pip install opencv_contrib_python
python -m pip install inky[rpi]==1.5.0
python -m pip install pillow

Install Large AI Models

• Refer to the open-source example by Vito Plantamura on GitHub: https://github.com/vitoplantamura/OnnxStream
• Go to the directory you created earlier and download and install XNNPACK

cd  ~/PaperPiAI
git clone https://github.com/google/XNNPACK.git
cd XNNPACK
git checkout 1c8ee1b68f3a3e0847ec3c53c186c5909fa3fbd3
mkdir build
cd build
cmake -DXNNPACK_BUILD_TESTS=OFF -DXNNPACK_BUILD_BENCHMARKS=OFF ..    # Here you need to download some corresponding repositories from github
cmake --build . --config Release    # The zero 2W here compiles for more than three hours 

• Go to the directory you created earlier and download and install OnnxStream

cd  ~/PaperPiAI
git clone https://github.com/vitoplantamura/OnnxStream.git
cd OnnxStream
cd src
mkdir build
cd build
cmake -DMAX_SPEED=ON -DOS_LLM=OFF -DOS_CUDA=OFF -DXNNPACK_DIR="${INSTALL_DIR}/XNNPACK" ..      # Here you need to download some corresponding repositories from github
cmake --build . --config Release      # Here zero 2W compilation takes about ten minutes 

• Download large model resources

cd  ~/PaperPiAI
mkdir models
cd models
git clone --depth=1 https://huggingface.co/vitoplantamura/stable-diffusion-xl-turbo-1.0-anyshape-onnxstream
#Download about 9G

AI Model Usage

Increase Swap Size for Compilation

• The default swap size of zero is 512, which is easy to lose WiFi and freeze when generating and parsing images
• Enter the command

sudo nano /etc/dphys-swapfile

• Change the value of conc_swapsize to 1024

• Press Ctrl + O to save, Ctrl + C to exit
• Restart dphys-swapfile

sudo /etc/init.d/dphys-swapfile restart

Self-built Models

• Download our package

wget https://files.waveshare.com/wiki/RPi_Zero_PhotoPainter/Demo/RPi_Zero_PhotoPainter.zip
unzip -o RPi_Zero_PhotoPainter.zip -d RPi_Zero_PhotoPainter
cd RPi_Zero_PhotoPainter
cp -r Waveshare_E-Paper ~/PaperPiAI     # The path for building large models before with PaperPiAI
cd cd ~

• Run a large model

cd PaperPiAI 
python Waveshare_E-Paper/generate_picture.py --steps 1  # Generate images using random instructions in the default configuration
python3 Waveshare_E-Paper/display_picture.py -m epd7in3e

Download Configured System

• Run a large model

cd PaperPiAI 
python Waveshare_E-Paper/generate_picture.py --steps 1  # Generate images using random instructions in the default configuration
python3 Waveshare_E-Paper/display_picture.py -m epd7in3e

Command Parsing

generate_picture.py

• Default output path (if you want to modify it, you need to put the directory at the end)

PaperPiAI/output_dir

• Default JSON retrieval (for image generation prompts)

Waveshare_E-Paper/flowers.json

• Other

--prompt     Use the prompt directly, overriding the prompt file (empty by default)
--prompts    Prompt the configuration file path (default Waveshare_E-Paper/flowers.json)
--seed	      Random seed reproduction
--steps      Number of generation steps (Higher = better quality but slower, default 5, recommended 1 step)
                 zero2w: 1 step takes about 34 minutes; 5 steps take about 1.5 hours
                 Pi5: 1 step takes about 5 minutes; 5 steps takes about 34 minutes
--width      Image width (default 800)
--height     Image height (default 480)
--sd         Stable diffusion binary path (not recommended to modify)
--model      Stable diffusion model path (not recommended to modify)
--overwrite  Overwrite existing output files
--no-copy    Do not copy to shared files

display_picture.py

• Default image input path

PaperPiAI/output_dir/output.png

• Default cropped image input path

PaperPiAI/output_dir
#The output name is:  processed_TIMESTAMP.png

• Other:

-o  Output path and name
-p  Portrait mode (forced rotation 90°)
-c  Center cropping, not smart cropping
-r  Only resize the image without cropping
-s  Simulated display, no actual output
-m  Display model (e.g., epd7in3e)
--skip-fit  If the image size matches the display perfectly, skip the adjustment

Use image data at the specified location (place the image position at the end of the command, no need to input the corresponding operation command)

python3 Waveshare_E-Paper/display_picture.py -m epd7in3e ./output_dir/processed_20250729_122744.png

Script Usage

• The command for generating images above is separate from the e-Paper screen refresh command. We have to wait for the image to be generated before refreshing the image, which is very inconvenient. Therefore, we provide a script code to automatically complete this process
• cron_flower.log Output English logs
• cron_flower-CN.log Output Chinese logs

#Download the demo (skip if already downloaded)
wget https://files.waveshare.com/wiki/RPi_Zero_PhotoPainter/Demo/RPi_Zero_PhotoPainter.zip
unzip -o RPi_Zero_PhotoPainter.zip -d RPi_Zero_PhotoPainter

#Enter the script directory
cd RPi_Zero_PhotoPainter/script_code
cp -r cron_flower.sh ~/PaperPiAI    # PaperPiAI is the path to building large models
#If you want to display Chinese logs, please transmit the following script
# cp -r cron_flower-CN.sh ~/PaperPiAI

# Enter the project directory (modify according to your own large model folder)
cd /home/pi/PaperPiAI

• You must add executable permissions to the script, otherwise it cannot run:

chmod +x /home/pi/PaperPiAI/cron_flower.sh

• Run script file

./cron_flower.sh

• Logs will be printed normally on the terminal, and the logs will also be saved in the /home/pi/PaperPiAI/image_update.log file. It is recommended to delete them regularly

> /home/pi/PaperPiAI/image_update.log

• If you are using cron_flower-CN.sh, the logs will be saved in the /home/pi/PaperPiAI/cron_flower.log file
• The delete command is

> /home/pi/PaperPiAI/cron_flower.log

Scheduled Refresh

• Make sure you have completed the above steps for installing the large model, demos, and scripts
• Enter the command

crontab -e

• If it's the first time entering, it will enter the compiler selection page; it is recommended to choose the nano editor; enter 1, then press Enter

Select an editor.  To change later, run 'select-editor'.
  1. /bin/nano        <---- easiest
  2. /usr/bin/vim.basic
  3. /usr/bin/vim.tiny
  4. /bin/ed

Choose 1-4 [1]: 

• Move the cursor to the end, then enter the scheduled execution command

0 8 * * * /home/pi/PaperPiAI/cron_flower.sh >> /home/pi/PaperPiAI/cron_log.log 2>&1

• Command parsing (which can be understood as the alarm setting of a mobile phone, presenting a similar effect)

0 8 * * *
  0: Minute
  8: Hour
  * * *: Year Month Day
/home/pi/PaperPiAI/cron_flower.sh
  Script path
>>
  Log append output → Each time the script runs, the log will be appended to the end of cron_log.log, and the file will continue to grow with each run;
/home/pi/PaperPiAI/cron_log.log
  The path where logs are saved
2>&1
  Redirect error logs to the same file as well, mixing them with normal logs;

• It is recommended to regularly clear the corresponding log files, the command is as follows

> /home/pi/PaperPiAI/cron_log.log

Time Zone Settings

• The system defaults to the time zone of 0, which is the time based on the Prime Meridian (0-degree longitude)
• You can set it using the following command, taking China as an example
• Enter the command

sudo dpkg-reconfigure tzdata

• Select Asia, then choose Shanghai or Beijing

• The following prompt will pop up after success

Demo Usage

e-Paper Driver

Hardware Connection

Please note that the PWR pin of this circuit is different from the PWR pin used in our general demo. If using the program from the general demo, make sure to modify it accordingly

Enable SPI Interface

• Open the Raspberry Pi terminal and enter the following command to enter the configuration interface:
  

sudo raspi-config
Select Interfacing Options -> SPI -> Yes to enable the SPI interface

• Restart Raspberry Pi:
  

sudo reboot

• Check /boot/config.txt and you can see that 'dtparam=spi=on' has been written to
  

• To ensure that SPI is not occupied, it is recommended that other driver overlays be temporarily turned off. You can use ls /dev/spi* to check the SPI occupancy. The terminal output /dev/spidev0.0 and /dev/spidev0.1 indicates that the SPI situation is normal

Run C Demo

• Install lg library

#Open the Raspberry Pi terminal and run the following commands:
wget https://github.com/joan2937/lg/archive/master.zip
unzip master.zip
cd lg-master
make
sudo make install
# For more information, please refer to the source code: https://github.com/gpiozero/lg

#Open the Raspberry Pi terminal and run the following commands:
sudo apt-get update
sudo apt install gpiod libgpiod-dev

#Open the Raspberry Pi terminal and run the following commands:
wget http://www.airspayce.com/mikem/bcm2835/bcm2835-1.71.tar.gz
tar zxvf bcm2835-1.71.tar.gz 
cd bcm2835-1.71/
sudo ./configure && sudo make && sudo make check && sudo make install
# For more information, please refer to the official website: http://www.airspayce.com/mikem/bcm2835/

#Open the Raspberry Pi terminal and run the following commands:
sudo apt-get install wiringpi
#For Raspberry Pi systems after May 2019 (those earlier may not require execution), an upgrade may be necessary:
wget https://files.waveshare.com/wiki/common/wiringpi-latest.deb
sudo dpkg -i wiringpi-latest.deb
gpio -v
# Run gpio -v and version 2.52 will appear. If it does not appear, there is an installation error.

#Bullseye branch system uses the following command:
git clone https://github.com/WiringPi/WiringPi
cd WiringPi
./build
gpio -v
# Run gpio -v and version 2.60 will appear. If it does not appear, there is an installation error.

• Download the demo (you can skip it if you have already downloaded it)

wget https://files.waveshare.com/wiki/RPi_Zero_PhotoPainter/Demo/RPi_Zero_PhotoPainter.zip
unzip -o RPi_Zero_PhotoPainter.zip -d RPi_Zero_PhotoPainter
cd RPi_Zero_PhotoPainter/7in3_e-Paper_E

• Compiler program (Note: -j4 means using 4 threads for compilation, the number can be modified by yourself)

# At this point in RPi_Zero_PhotoPainter/7in3_e-Paper_E location
cd c
sudo make clean
sudo make -j4

• Run the demo

sudo ./epd

Run Python Demo

• Install function library

sudo apt-get update
sudo apt-get install python3-pip
sudo apt-get install python3-pil
sudo apt-get install python3-numpy
sudo pip3 install spidev

sudo apt-get update
sudo apt-get install python-pip
sudo apt-get install python-pil
sudo apt-get install python-numpy
sudo pip install spidev

sudo apt-get update
# python3
sudo apt install python3-gpiozero
# python2
sudo apt install python-gpiozero

• Download the demo (you can skip it if you have already downloaded it)

wget https://files.waveshare.com/wiki/RPi_Zero_PhotoPainter/Demo/RPi_Zero_PhotoPainter.zip
unzip -o RPi_Zero_PhotoPainter.zip -d RPi_Zero_PhotoPainter
cd RPi_Zero_PhotoPainter/7in3_e-Paper_E

• Run the demo

# Ensure in RPi_Zero_PhotoPainter/7in3_e-Paper_E location
cd python/examples/
python3 epd_7in3e_test.py