- Python package for streaming video from multiple cameras to disk.
- Features real-time hardware-accelerated video compression and debayering using FFmpeg.
- (Optional) AMD or Nvidia GPU for hardware encoding (see https://developer.nvidia.com/video-encode-decode-gpu-support-matrix)
- Basler and/or FLIR machine vision camera(s)
- Windows or Linux PC
- Arduino/Teensy microcontroller for syncing cameras
- Update graphics drivers
- Create and activate a new Python 3.7 Anaconda environment:
conda create -n campy python=3.7 imageio-ffmpeg matplotlib -c conda-forge
conda activate campy
pip install -U setuptools
- Install camera software
-
If using Basler cameras, install Pylon software:
- Install Basler Pylon with Developer options
- Install pypylon: Windows:
pip install pypylonLinux:
conda install swig git clone https://github.com/basler/pypylon.git python ./pypylon/setup.py install -
If using FLIR cameras:
- Download and install Spinnaker SDK and SpinView software from FLIR's website: https://www.flir.com/support-center/iis/machine-vision/downloads/spinnaker-sdk-and-firmware-download/
- Manually install binary wheel for PySpin (included in the Spinnaker download) E.g. for Python 3.7 on Windows amd64 system, install "spinnaker_python-2.3.0.77-cp37-cp37m-win_amd64.whl"
pip3 install <wheel>
- Clone or download campy to local folder:
git clone https://github.com/ksseverson57/campy.git
- Finally, install campy and its dependencies (see setup.py) by navigating to campy folder:
pip install -e .
- For Basler cameras, use the Pylon Viewer to save your '.pfs' camera settings file. Examples are included in campy/cameras/basler/settings.
- Edit the config.yaml file to fit your system and recording configuration.
- Several example config files are located in campy/configs.
- For help setting config parameters:
campy-acquire --help
Campy's trigger module supports Arduino and Teensy microcontrollers:
- Download Arduino IDE (https://www.arduino.cc/en/software). If using Teensy, install Teensyduino (https://www.pjrc.com/teensy/teensyduino.html).
- Connect your microcontroller and note its port number (e.g. "COM3" on Windows or "/dev/ttyACM0" on Linux).
- In your config.yaml, configure:
startArduino: True
digitalPins: [<pin IDs>] # e.g. [0,1,2]
serialPort: "<port>" # e.g. "COM3" or "/dev/ttyACM0"
- Open and upload "trigger.ino" file (in campy/trigger folder) to your board. Make sure serial monitor is closed while using pyserial connection.
- Campy will synchronously trigger the cameras once acquisition begins.
campy-acquire ./configs/campy_config.yaml
- Campy will stop automatically after set recording time (e.g. 1 hour):
recTimeInSeconds: 3600
- To manually end, press Ctrl^C. Wait until campy exits.
- Three files, "frametimes.mat", "frametimes.npy", and "metadata.csv", will be saved along with the video file in each camera folder containing timestamps, frame numbers, and other recording metadata.
- To debug broken pipe error, include this in config.yaml:
ffmpegLogLevel: "warning"
- Use the command "ffmpeg" to check enabled packages
- On Linux, you may need to compile your own ffmpeg binary to enable encoders:
- Nvidia:
git clone https://git.videolan.org/git/ffmpeg/nv-codec-headers.git
- Intel:
sudo apt-get install libva-dev libmfx-dev libx264-dev libx264-dev libnuma-dev
- AMD:
git clone https://github.com/GPUOpen-LibrariesAndSDKs/AMF.git
sudo cp -r ./AMF/amf/public/include/core /usr/include/AMF
sudo cp -r ./AMF/amf/public/include/components /usr/include/AMF
- Compile ffmpeg:
git clone https://git.ffmpeg.org/ffmpeg.git
cd ffmpeg
sudo ./configure \
--enable-cuda --enable-cuvid --enable-nvdec --enable-nvenc --enable-nonfree --enable-libnpp \
--extra-cflags=-I/usr/local/cuda/include --extra-ldflags=-L/usr/local/cuda/lib64 \
--enable-gpl --enable-libx264 --enable-libx265 --enable-libmfx \
--enable-amf
sudo make -j -s
sudo cp -r ./ffmpeg /usr/bin
- Include in config.yaml:
ffmpegPath: "/usr/bin/ffmpeg"
Written by Kyle Severson with contributions from Diego Aldarondo and Iris Odstrcil (2019-2021).
Special thanks to Tim Dunn, David Hildebrand, Vincent Prevosto, Manuel Levy, and Paul Thompson for helpful comments.
MIT License