You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I added a C++ version of the basic algorithm using OpenCV.
A DVS is a camera-like sensor which emits per-pixel asynchronous Address-Event Representation (AER) packets. These are originated due to an intensity level change has reached a certain threshold.
This project emulates this behaviour using commodity hardware, that is a regular PC and camera. Modifications to the default behaviour where added, mainly: encoding of values using time, local inhibition and dynamic thresholds.
Fred Rotbart (https://github.com/fcr) has been kind enough to port to Python 3 and checked compatibility with MacOS Sierra, plus code clean-up!
Requirements:
Cython
Numpy
OpenCV (with Python bindings)
If using as an application, before using, compile the Cython modules by running
python setup.py build_ext --inplace
in the folder where pyDVS is installed.
Otherwise, it can be simply installed as a library by running
pip install .
in the folder where pyDVS is installed.
Files:
pydvs/:
generate_spikes.pyx - Cython backend library. Contains functions to generate spikes from a video feed.
external_dvs_emulator_device*.py - sPyNNaker class wrapper, it's compatible with OpenCV VideoCapture devices and VirtualCam.
virtual_cam.py - A camera emulator that feeds on images located in the users computer. It keeps images moving based on eye movements (e.g. micro saccades and saccades); there's a very simple attention model for the saccade mode.
spynnaker_test.py - Demonstration script for sPyNNaker compatible frontend.
stand_alone*.py - Demonstration scripts for the Cython backend, they come in single- and multi-threaded versions.
virtual_cam_test.py - Demonstration script for the VirtualCam class, it requires images in a directory ("./mnist" by default)
The multithreaded versions require either Python 3.4 and above or Windows
About
An emulation of a Dynamic Visual Sensor using Cython/Numpy for calculations.