Data from images - photographs and video

Hi everyone, especially @srudinac @Inske @dhaval

As mentioned in today’s Partner teleconference, Bradford Council has commissioned some drone footage of several large road junctions in the Bradford area.

It is hoped that using machine learning, image recognition etc. that traffic movement statistics can be obtained.
Much of this work is currently done manually with several people on location making direct observations and counting.
There might be other data that can be derived, perhaps regarding ‘pinch points’ where there is more possibility of collision, or areas where there are frequent inappropriate maneuvers - dangerous driving and there’s a lot of that in Bradford!

Here is a sample of what might be possible:

Nice project @sydsimpson !

I look forward to hearing more about this in Bergen

@Inske @h.niesing hello Inske and Hugo. I’m working on the CCTV Refuse vehicle imagery that I showed in Bergen and I think @Jaccobrouwer may have mentioned that Amsterdam have developed some software to ‘blur’ faces and vehicle number plates. Are either of you aware of the software? Or any other free software that might accomplish this? Before I supply the data to anyone i really do need to ‘blur’ faces and vehicle number plates.

@klantto Hello Kim is this anything that you are aware of?

Hi @Adrian, Hi @TomOS,

we are really interessted in the results from the Workshop you had after the partnership Meeting in Bergen. Is there anything you can report from that?

Best regards
Henning

Hello Henning,
I hope you are keeping well.

I think Microsoft were interested in the drone footage and I believe that they were going to provide some feedback regarding how they could obtain data from the imagery. @sydsimpson @phezoj @chrise68 was there anything else that was relevant to the drone footage?

Hi Adrian,

Great post! I think this could be a great common component to collaborate around.

I did some research and found this existing open source tool

• https://github.com/understand-ai/anonymizer

Some other forums pointed me to

• https://github.com/openalpr/openalpr - which apparently can be repurposed to include blurring

And also found these two, which are not open source

• https://www.sighthound.com/products/redactor
• https://orbitgt.com/products/3dm-blur-erase/

– Edit –

Additional option shared by @pjppauwels in later comment

• Mapillary blurring tool

Hi @Adrian, I’m unfortunately not aware which software Amsterdam used to blur the faces. @h.niesing Hugo, do you know?

@sydsimpson I just came across this repo that Maarten (another PhD student under the supervision of Marcel and Stevan) shared:

• https://github.com/maartensukel/video-to-traffic-flow-information/

I thought this might help you with analyzing your drone footage.

Yes I think it is Bert or Marten from the development team.

But Iwill check tomorrow and get back to you

Gr Hugo

Thank you @claus I will keep you informed

Hi all

@pjppauwels says hi and he shared following information to be added to @claus’s desk research

Also, I came across this video on LinkedIn demoing the repo shared by @Inske

@timvanachte @pjppauwels this looks fantastic how can we implement/test this? @sydsimpson @phezoj @dhaval @Bhupesh

Due to disruption from the Covid pandemic, I’ve only just managed to get the Drone footage, Data Protection Impact Assessment (DPIA) approved (a new Data Protection Officer definitely helped the process). I’m just getting the Data Sharing agreement approved by our Data Protection Officer; which will need signing by each organisation that would like a copy of the videos.

Please let me know if you would like copies of the videos.

@sydsimpson @phezoj @dhaval @Bhupesh @timvanachte @pjppauwels