Model taked from: https://github.com/nianticlabs/monodepth2

Colab testing environment: https://colab.research.google.com/drive/1TXeI7PwdrbFNWTkw7RTdptvMD4-pAV2V?usp=sharing

Model: Raspberry Pi Zero W V1.1

Architecture: armv6 32bits

• First Install Raspbian

  • Reset a MicroSD and install Raspian OS on it

• Update Linux package handler

  • sudo apt-get update
    

  • sudo apt-get upgrade
    

• Enable camera

  • sudo raspi-config
    

  • Select Interfaces Option

  

  • Legacy camera

  

  • Enable

• Install python 3.7

  • Follow steps in: https://programmerclick.com/article/3973980601/

• Update pip

  • python -m pip install --user --upgrade pip
    

  • Check version:

  • python -m pip --version
    

• Install torch

  • Git Raspberry Pi Zero V1 compatible torch versions: https://github.com/nmilosev/pytorch-arm-builds
  • Steps:
    • Install necessary libraries:

    •   sudo apt install libopenblas-dev libblas-dev m4 cmake python3-dev python3-yaml python3-setuptools python3-wheel python3-pillow python3-numpy
      

    • Download versions in the git:

    •   wget https://github.com/nmilosev/pytorch-arm-builds/archive/refs/heads/master.zip
      

    •   unzip master.zip
      

    •   cd pytorch-arm-builds-master/
      

    • Install armv6 compatible torch and torchvision versions:

    •   python -m pip install --user ./torch-1.1.0-cp37-cp37m-linux_armv6l.whl ./torchvision-0.3.0-cp37-cp37m-linux_armv6l.whl
      

    • Fix Pillow error:

      •   python -m pip uninstall pillow
        

      •   python -m pip install --user "pillow<7"
        

    • Veryfy by running:

    •   python inference.py <image path>
      

• Install PiCamera

  • python -m pip install --user picamera
    

• Download the project scripts from this git on your Raspberry Pi and run it