It's funny that this is released the same day as a 56 minute overhead drone video of the Swindon magic roundabout, one of the UK's most iconic road junctions: https://youtu.be/CzQunCuQCJY?si=X1nbcxiWOCcnm8eO
An interesting effect of counting at roundabouts is that MHT/basic inertial movement models tend to work quite well, because the vehicles aren't stopping and starting, and they tend to not bunch up as much as at a red light
Seems to be made for USA as only cars are counted. In Europe u usually want to distinguish between cars bigger cars, bicycles and pedestrians. You also need to blurr everything not related to cars and you can't just collect all video data in a central space. Or need lower res cameras.
<i>I think a combined detector/associator network is the next step.</i>
Repentance for using MHT to accelerate vehicle interconnectedness might consist of V2I.
Audi vehicles with 802.11p V2I capabilities have an information sharing agreement with Traffic Technology Services in certain municipalities that alert a driver of a vehicle when a red light will turn green.
You may not reverse engineer, decompile, or disassemble the SOFTWARE PRODUCT, except and only to the extent that such activity is expressly permitted by applicable law notwithstanding this limitation.
...
COPYRIGHT
All title and copyrights in and to the SOFTWARE PRODUCT (including but not limited to any images, photographs, clipart, libraries, and examples incorporated into the SOFTWARE PRODUCT), the accompanying printed materials, and any copies of the SOFTWARE PRODUCT are owned by the Author of this Software. The SOFTWARE PRODUCT is protected by copyright laws and international treaty provisions. Therefore, you must treat the SOFTWARE PRODUCT like any other copyrighted material. The licensed users or licensed company can use all functions, example, templates, clipart, libraries and symbols in the SOFTWARE PRODUCT to create new diagrams and distribute the diagrams.
It's funny that this is released the same day as a 56 minute overhead drone video of the Swindon magic roundabout, one of the UK's most iconic road junctions: https://youtu.be/CzQunCuQCJY?si=X1nbcxiWOCcnm8eO
An interesting effect of counting at roundabouts is that MHT/basic inertial movement models tend to work quite well, because the vehicles aren't stopping and starting, and they tend to not bunch up as much as at a red light
Seems to be made for USA as only cars are counted. In Europe u usually want to distinguish between cars bigger cars, bicycles and pedestrians. You also need to blurr everything not related to cars and you can't just collect all video data in a central space. Or need lower res cameras.
Very cool, thanks for sharing. How did you train it? Just manually labeling the data?
I hired some workers in Bangladesh
<i>I think a combined detector/associator network is the next step.</i>
Repentance for using MHT to accelerate vehicle interconnectedness might consist of V2I.
Audi vehicles with 802.11p V2I capabilities have an information sharing agreement with Traffic Technology Services in certain municipalities that alert a driver of a vehicle when a red light will turn green.
https://arstechnica.com/cars/2018/03/cars-that-talk-to-traff...
Just tossed it over to my GIS and metro-planning friends and they all reacted with "O-hoooo, what a good news" .
Kudos!
Excellent, I'm going to post links to installers to make it somewhat more convenient, just haven't gotten to that yet
I made an Android app in this space a few years back. It's helped a few people but I never got a grip on the market or understanding of the space.
www.roadcount.com
Looks like it might be using background subtraction, is that right?
Its running EfficientDet on the phone using Tensorflow light.
[stub for offtopicness]
Redacted/updated repo with BSD 3-Clause license:
https://github.com/asfarley/vtc_opensource
Sorry for the confusion
Ok, we've changed the URL to that from https://github.com/asfarley/vtc_lfs above. Thanks!
your github repo link is broken, perhaps you forgot to unprivate it.
Whoops, thank you. Fixed now.
404?
I accidentally linked the wrong repo (it was original/unredacted with API keys, etc). The correct repo is: https://github.com/asfarley/vtc_opensource
Assume someone grabbed the API keys. Change them as soon as possible.
Changed now. Thanks!
From the EULA.txt [0]:
""" ...
DESCRIPTION OF OTHER RIGHTS AND LIMITATIONS
You may not reverse engineer, decompile, or disassemble the SOFTWARE PRODUCT, except and only to the extent that such activity is expressly permitted by applicable law notwithstanding this limitation.
...
COPYRIGHT
All title and copyrights in and to the SOFTWARE PRODUCT (including but not limited to any images, photographs, clipart, libraries, and examples incorporated into the SOFTWARE PRODUCT), the accompanying printed materials, and any copies of the SOFTWARE PRODUCT are owned by the Author of this Software. The SOFTWARE PRODUCT is protected by copyright laws and international treaty provisions. Therefore, you must treat the SOFTWARE PRODUCT like any other copyrighted material. The licensed users or licensed company can use all functions, example, templates, clipart, libraries and symbols in the SOFTWARE PRODUCT to create new diagrams and distribute the diagrams.
... """
At the bottom of the README:
""" © Roadometry 2018 """
Maybe I missed something but this seems pretty closed source to me.
"Source available" would be a more appropriate term for this project.
[0] https://github.com/asfarley/vtc_lfs/blob/master/EULA.txt
I botched the open-sourcing, I intended to post this one: https://github.com/asfarley/vtc_opensource
I've updated the license to BSD 3-Clause and deleted the EULA.txt.
In that repo, the default branch 'main' has only a licence file. Perhaps you meant to make the 'master' branch the default?
Yes, done now