Processing Video Stream using Doubango's Ultimate ALPR SDK

[Abd Shomad]

Dear Group Members, 

Pardon me for the newbie question.

I'm watching below result from the Doubango AI Youtube channel and the result looks impressive. 

https://youtu.be/xQO7ABHTg1w

Context: I am working as a system integrator implementing ALPR solution in already 4 cities, and now considering to replace our current ALPR implementation (O*** ALPR) and our future implementation with other ALPR solutions.

In order to convince, both technical and management team, I needed to be able to compare our video stream, processed by both our existing ALPR and Doubango's Ultimate ALPR side by side. 

I have downloaded the SDK and trying the recognizer.exe and recognizer.bat on my windows and linux machine. However, I could not find the executable/sample code/documentation on how to process video stream and produce plate recognition like above.

Could someone kindly enough direct me to the correct source/samples/executable/documentation to produce like the above video using Doubango's Ultimate APLR SDK? 

NOTE: I have read explanation on this group that the SDK will mark "*" for some characters during evaluation, it is OK for me. 

Best Regards, Abd Shomad 

[Mamadou DIOP]

Hi,

The samples in our Github repo only accept images as input. The code to read from a camera our rtsp stream is up to the developer and very easy to write. For example, you have sample code using OpenCV at https://www.doubango.org/SDKs/anpr/docs/Getting_started.html#opencv.

It’s very easy to reproduce the result from the video using OpenCV and off course you have to write code. We haven’t shared our code because it’s based on CompV (https://github.com/DoubangoTelecom/compv) not OpenCV. More at https://groups.google.com/g/doubango-ai/c/6SMsK5Us9bQ/m/JypzoPk2DgAJ. 

Regardless the backend you use to decode the video you’ll get the same accuracy. You can also use VLC Player to grab random images from the video and test them against recogniser.exe or https://www.doubango.org/webapps/alpr/. I know this is not what you’re asking for but it’ll give you an idea about the accuracy. The video was generated using v3.0-beta, the current version is v3.3 and way more accurate.

This said, I don’t think putting videos side-by-side is the best way to compare products. Build a 500+ dataset and compute a recall and precision score for each product.

Regards,

--
You received this message because you are subscribed to the Google Groups "doubango-ai" group.
To unsubscribe from this group and stop receiving emails from it, send an email to doubango-ai...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/doubango-ai/dbbcc7a7-552f-4cd6-bc06-99772cfe1fban%40googlegroups.com.