readme.md

Nov 2018

Building object tracker workspace for ping pong ball.

Multiple utilities (alpha version) for recording, viewing, and characterizing video data from a consumer webcam. These are used to build and organize training/evaluation data which can be used to tune object tracker algo's for a desired task and time/accuracy characteristics.

The directory data/ in this repository corresponds to the ppd_data repository here: https://github.com/sutt/ppd_data

See documentation for the utility scripts here: https://github.com/sutt/ppd/tree/master/docs

Quickstart

git clone https://github.com/sutt/ppd.git

cd ppd
rm data/ -r
git clone https://github.com/sutt/ppd_data.git data

cd test/
pytest -vv     (some tests need a webcam available and will use it)

To view a simple example:

python guiview.py --file data/proc/tmp/demo-agenda/output4.proc2.avi --showscoring

(use gui ('Tracking' to On) to turn on tracking and visually evaluate performance)

requirements:

python 2.7.13

numpy 1.14.5+mkl
opencv-contrib-python 3.2.0.8
matplotlib  2.0.1
pytest      3.3.2
imutils     0.4.3
Tkinter     revision 81008

(windows supplementary .dll's)
[not sure which one right now]
openh264-1.4.0-win32msvc.dll
openh264-1.4.0-win64msvc.dll
opencv_ffmpeg320_64.dll
opencv_ffmpeg320.dll

(optional:)
jupyter
Flask
any others?