a-lohia / vsim

VSim Differential Motion Simulator

Overview

VSim is a differential motion simulator designed to create simple toy videos to test with potential solutions for the ego-motion problem. Using VSim you can create videos of a square moving in a custom path created by your mouse on a constantly moving background.

input.mp4

input_REF.mp4

Installation

Clone this repository. Then run pip install -r requirements.txt to install. If you want to convert these videos to DVS, install v2e sim by cloning the v2e repository as well

Important

VSim requires ffmpeg to be installed on your system

Instructions

Parameters

Parameter	Description	Implemented
save_dir	Path to where you want to save the test data	Yes
ref_save_dir	Path to where you want to save the reference test data	Yes
height	Height of the test data (in pixels)	Yes
width	Width of the test data (in pixels)	Yes
frames	Number of frames	Yes
fps	Frames per second of the test data	Yes
col_width	Width of the columns in the moving background (in pixels)	Yes
object_radius	Radius of the moving object (in pixels)	Yes

Instructions to use

1. Add the parameters to your params.yaml file
   • save_dir is the path to where you want to save the test data USE .mp4 or it will not work
   • ref_save_dir is the path to where you want to save the reference test data (for use as "ground truth" in the future) USE .mp4 or it will not work
   • height is the height of the test data (in pixels)
   • width is the width of the test data (in pixels)
   • fps is the frames per second of the test data
   • col_width is the width of the columns in the moving background (in pixels)
   • object_radius is the radius of the moving object (in pixels)
   • fill is the amount of fill (-1 is filled, 0 is border) > 0 is thickness
2. call create_data.create_data("path_to_params.yaml") to create the test data (This is in manual path mode, so you will have to manually create the path with your mouse)
   • You can enter a discrete path instead of the create-your-own by assembling your own function and following the steps in test.py

Credits

Written by Arya Lohia for the IRIS project at Parsa Research Lab at GMU

License

MIT License