README.md

In dsidavis/VideoEvents: Process videos to get motion event start, end and duration

A simple way to run DVR-scan to analyze videos for motion events and get the results back into R. We run the pythong code for DVR-scan from R.

DVR-scan: https://dvr-scan.readthedocs.io/en/latest/

To install the package

devtools::install_github("dsidavis/VideoEvents")

Or git clone the repository and install it locally - from within in R

install.packages("VideoEvents", repos = ".")

or on the shell command line

R CMD INSTALL VideoEvents

Load the package to get the 3 functions:

library(VideoEvents)

Process a single video. We'll use the one of the rug that is included in the package

f = system.file("videos/Other/54.mp4", package = "VideoEvents")
dvrScan(f)

This assumes the dvr-scan command is available in your path as a regular executable. I don't have it installed so I have to call python3 dvr-scan.py and I have to specify where the dvr-scan.py file is so I call the R dvrScan() function as

e = dvrScan(f, exec = "python3 ../DVR-Scan-1.0.1/dvr-scan.py")

Given the result, we can convert it to a data frame with

e = events2DataFrame(e, f)

This video has no events, so we end up with a row with NAs.

To process all the videos in a directory and its subdirectories and convert the events to a data frame all in one step, use the processVideos() function:

i = processVideos("inst")

You can run this using more than one CPU/core so that they process each video in parallel. You specify how many cores to use with num.cores = n, e.g.,

i = processVideos("inst", mc.cores = 3

dsidavis/VideoEvents documentation built on June 29, 2019, 12:12 a.m.