Dlib: plate detection

[MyraBaba]

Hi,

Is it a good idea to use dlib for vehicle plate detection and recognition like openalpr ? Is there anybody has experience ? What could be the best starting point if advised?

Best

[edubois]

I think so, but I would make a Generative Adversial Network for that, and then extract the detector.

[davisking]

Is there a paper that describes that kind of technique?

[MyraBaba]

This is what I found :
Adversarial Generation of Training Examples: Applications to Moving Vehicle License Plate Recognition

https://arxiv.org/pdf/1707.03124.pdf

[davisking]

I think this isn't a great idea. You would be better off training a normal
detector to find license plates, or using a non-ML based text detector like
maximally stable extremal regions or the stroke width transform.

[edubois]

I would generate fake examples by means of the GAN, and insert true / false examples from another labelized set when training the discriminator. This semi supervised method would give you more examples. But making a GAN is a bit tricky.

[edubois]

We'd love to have a Conditionnal GAN in a future release of dlib :)

[MyraBaba]

I will test with normal detector and let you know the results.

[AKSHAYUBHAT]

The visual genome dataset contains few hundred images with bounding box for license plates.

[dlib-issue-bot]

Warning: this issue has been inactive for 191 days and will be automatically closed on 2018-09-07 if there is no further activity.

If you are waiting for a response but haven't received one it's likely your question is somehow inappropriate. E.g. you didn't follow the issue submission instructions, or your question is easily answerable by reading the FAQ, dlib's documentation, or a Google search.

[dlib-issue-bot]

Notice: this issue has been closed because it has been inactive for 195 days. You may reopen this issue if it has been closed in error.