Darknet/yolo/openCV looks promising. Thanks
See for example: Automatic License Plate Detection & Recognition using deep learning | by Achraf KHAZRI | Towards Data Science
Which deploys differing technologies to perform 3 distinct steps …
- Localize
- Segment
- Recognize
1 Like
Ps you still haven’t told us much about this ‘requirement’ to extract (and fuse?) information from multiple input images.
Thank you very much @ai_curious
The problem is evolving and possibly one image will not have the necessary info and hence would need to be fused
As shared by @ai_curious , fusing images would not be part of YOLO capabilities.
When you say “fuse”, how will this ‘mechanically’ work? is this a real time scenario? will you have 2 cams capture video or taking pictures from different angles? If an image will not have the necessary information, do you mean that the license plate is attached to an object (car?) and you need to take pictures of that object from several angles (front, side, back) to capture more than the license plate information?
Not sure Juan… am just trying to prepare by getting information on which models to start investigating
I would say, to get information of models, I would want to have a clear definition of the specs. Without clear specs, any model can seem to work or not. As the saying goes: with no clear objective, any direction is fine. 
1 Like
For someone learning from experts, any information is useful information 
Thanks
+1 on solidifying business or functional requirements before going too far with technology. In my experience, if the problem cannot be crisply described in human language, it is premature to try to codify into computer language.
Found this while researching a different but related learner inquiry…
(Though I’m still not completely clear on the functional requirements)