Would you like to see person detection as a feature in ZM? Add to Bounty.

Forum for questions and support relating to the 1.31.x releases only.
Post Reply
User avatar
asker
Posts: 1049
Joined: Sun Mar 01, 2015 12:12 pm

Would you like to see person detection as a feature in ZM? Add to Bounty.

Post by asker » Mon Jul 02, 2018 3:55 pm

(posted in 1.31.x because not sure how many people actually ever look at Feature Requests)

Would you like to see person detection as a feature in ZM?

I don't believe its hard, but we need to make it worth the time for the ZM folks. I've set up a GitHub issue with a Bounty. Other mechanisms today like integrating machine learning as post processing scripts are not ideal (a) they are very slow b) they only kick in once an event is detected. We need this to be integrated into ZM's core.

I don't think adding machine learning based detection (CNNs like TinyYOLO, FastRNNs etc) are a good idea - they are still too slow to work properly in non GPU machines, hence my suggestion to add OpenCV's HOG mechanism. Not as accurate, but much faster. It will be a big step forward from blob detection.


If you think its useful to you too, please feel free to add to the bounty. Lets make it worth it for the devs to prioritize.

https://github.com/ZoneMinder/zoneminder/issues/2138
Bounty Link: https://www.bountysource.com/issues/603 ... ate-option
--
My collection of ZoneMinder learnings:
https://wiki.zoneminder.com/Various_ZM_thoughts

nrbell
Posts: 11
Joined: Mon Jun 04, 2018 1:14 pm

Re: Would you like to see person detection as a feature in ZM? Add to Bounty.

Post by nrbell » Mon Jul 16, 2018 4:05 pm

I think that this is a great idea. Having a framework that would easily allow the integration of future OpenCV-based features would be a big step up for the ZoneMinder platform. I've contributed to the bounty.

rockedge
Posts: 1079
Joined: Fri Apr 04, 2014 1:46 pm
Location: Connecticut,USA
Contact:

Re: Would you like to see person detection as a feature in ZM? Add to Bounty.

Post by rockedge » Wed Jul 18, 2018 1:37 am

I have tested some routines and will look at this idea as well. I like the frame work idea and trying HOG

Post Reply

Who is online

Users browsing this forum: No registered users and 2 guests