PyAER

PyAER with Swig Bindings

Special thanks to iniVation for making this possible.

The project is in its Beta development stage, please submit an issue if you need our help.

Check out the documentation for more help!

Why PyAER?

iniVation has released DV, a new platform and an SDK for accessing and developing with event cameras.

For robotics projects, you may find rpg_dvs_ros when you use ROS.

So the natural question is: why PyAER?

Although there are large function overlaps between these frameworks, PyAER serves as an agile package that focus on fast development and extensibility. In fact, in some scenario, e.g. edge devices, PyAER is more user friendly than other alternatives.

Design Principle

Minimum installation effort.
Clean, simple, easy to manage.
Well documented, human-readable code.

Installation

Install libcaer dependency (RECOMMEND)
```
$ bash <(curl -s https://raw.githubusercontent.com/duguyue100/pyaer/master/install-libcaer.sh)
```
NOTE: For more information, see libcaer repo.

NOTE: From 0.1.0a18, we support eDVS, you will need to install libserialport so that the package can work properly, follow the building instructions from here. Currently, this support is not built into the release since we are not clear how useful is this feature. If you are interested, you can build the project from scratch.
Install pyaer from pypi (RECOMMEND)
```
$ pip install pyaer
```

Install from source

$ git clone https://github.com/duguyue100/pyaer.git
$ make install

Got a Linux? [Optional]

NOTE: The libcaer installation has taken care of this problem. However, if the problem persists, please try following.

libcaer relies on libusb based driver, you won't be able to access the camera unless fixing the udev rules. Refer to details at here.

$ bash <(curl -s https://raw.githubusercontent.com/duguyue100/pyaer/master/install-udev.sh)

Running Examples

The scripts folder provides some examples for you to play with:

dvs128-test: you need OpenCV to run this example, note that if you are on Mac, OpenCV's waitKey() function may cause delay of displaying frames.
dvs128-glumpy: you need glumpy package to run this example. glumpy is a fast visualization library based on OpenGL. We found it's very fast to render images. In our case, we use GLFW backend. If glumpy couldn't find your installed GLFW on your system, make sure you set the GLFW_LIBRARY variable to /your/glfw/library/path/libglfw.so.
dvs240-test: you need OpenCV to run this example.
dvs346-test: you need OpenCV to run this example.

More examples are coming...

Yeah, you need SWIG

You only need to read this section if you are planning to compile pyaer from source.

This repository uses SWIG to create Python bindings. And you will need to compile the latest SWIG from source. The reason is because current SWIG cannot handle some cases in libcaer, we made a modified SWIG for this purpose.

Install compilation dependency
```
$ sudo apt-get install automake
$ sudo apt-get install bison
```
There might be other dependencies for compiling SWIG

Compile SIWG

$ git clone https://github.com/duguyue100/swig
$ cd swig
$ ./autogen.sh
$ ./configure
$ make
$ sudo make install

For compiling SWIG with Python

$ ./configure --without-alllang --with-python=$(command -v python)

NOTE: If you are not compile the SWIG with system Python distribution, it won't link to the custom Python automatically.

You will need to configure LD_LIBRARY_PATH for swig running properly.

i.e.

LD_LIBRARY_PATH=$HOME/anaconda2/lib:$LD_LIBRARY_PATH swig

Limitations and Notes

Current status of the project is meant for single device use. Potentially, this library supports multiple devices at the same time by giving concrete device names and serial numbers. Supporting and testing for multiple devices setup is in long-term plan, but we are not working on this right now.
Once the data stream is open, the data will be streamed through USB connection at certain publishing frequency (e.g. 100Hz). This is a hardware configuration, therefore you couldn't drop event packets by putting software-level delay. You can either skip processing the coming packets by some conditions or implement a queuing system that can do a particular dynamic fetching.
It's recommended to implement a multi-processing or multi-threading program so that each process or thread only deals with one particular task. The fetching of the event packets may be very fast, your program may be delayed if you are not carefully coping with this fact.
DYNAP is generally supported. We are currently looking for the correct bias configuration mechanism so that it can easily support the use of the device. We have mapped some core functions that are essential to device configuration.

Contacts

Yuhuang Hu
Email: duguyue100@gmail.com

franciumlee / pyaer