Help needed to gather images from users
  • Hello SecuritySpy Users!

    We are planning to make SecuritySpy much smarter about motion detection and analysis of video data. To achieve this, we are working on a neural network algorithm that can intelligently analyse and categorise images and video. To enable us to do this effectively, we need to collect a large amount of images from video surveillance cameras to use as training data.

    In the latest version of SecuritySpy (4.2.5), there is a new option in the General Preferences called "Send anonymous images back to developer". When this is enabled, SecuritySpy will send images from your cameras back to us when it detects motion. We can then sort and categorise these images and use them as training data for the neural network we are developing.

    image

    For privacy, this option is off by default. We need as many users as possible to send us images, so if you don't mind doing so, please enable this option. Images are sent infrequently and are scaled down to a small size, so this will not impact your Internet connection. Apart from the content of the images themselves, no other personal data is sent with the images.

    Many thanks in advance to anyone who is willing to help us out with what will be a major improvement to SecuritySpy.

    Ben
  • Ben, could there be an option to select which cameras are involved?
    Would be happy to help supply images, but not necessarily from ALL cameras.
  • Hi Guy, that's a good idea, I didn't think of that! I'll see if I can add a choice of cameras and release this as a beta soon. Will let you know when done. Thanks.
  • OK I have now added the option to select specific cameras for this, in the latest beta version of SecuritySpy (currently 4.2.6b1). When you enable the option, a panel will pop up that allows you to choose which cameras to include.
  • Great. Will do tonight.
  • Really happy to see this. Two places where I have the most issues and I hope can be addressed with object detection:
    1. small motion, e.g. somebody distant moving towards/away from the camera. If I have the sensitivity too high, I get a lot of false positives, and if too low, no recording. I've taken to recording long sections before and after motion detection so I can get the "complete view." (I live next to an open lot/field that's marked with no trespassing signs and have occasional unsavory foot traffic to/from it. I'd like to get an eye on people sooner/later than my motion detection settings allow).
    2. shadows of trees swaying in the wind. Today the trees are easily masked, but obviously the shadows move throughout the day.
  • Hi @bp33 thanks for the input. The new features we are planning should totally eliminate number 2 but may not help with number 1. The moving object has to be a certain size in the frame before any motion-detection algorithm (even a neural network) can analyse it properly. Small and distant motion will always be ignored, as this is the right thing to do in virtually all cases.
  • A like option would be awesome in the forum! I have noticed the same issue as bp33. Send images enabled btw!

Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!