Face and license plate detection in equirectangular images

Hi all,

We’ve been experimenting with some basic detection methods to blur out sensitive information in our photos. For example cars and people.

The solution is not particularly elegant, but it works pretty well. The blurring logic covers entire cars and people as opposed to faces and license plates.

The biggest issue right now is equirectangular projections. Detection of objects works well for less distorted areas (center of photo) but objects in ‘stretched’ areas around the top and bottom of the image are regularly missed (as the models I’m using are designed for flat images).

Ideally I’d like to get it to a level comparable to Google Street View or Mapillary where less of the image is spoilt (e.g. not an entire car blurred), but identifiable info is consistently removed.

Does anyone have any links to work others working on this problem? Or suggestions?

Check out this thread: https://forum.mapillary.com/t/anonymize-blurring-faces-and-license-plates-before-upload/4230

Some of the tools mentioned:

Hi David,
This doesn’t answer your question but I wanted to share that Google’s own solution doesn’t scan the whole image for number plates it seems.

It doesn’t pick up old type number plates and number plates mounted higher up than usual.

2 examples below from a large steam rally capture. -

Old style number plate on the front of a steam engine.

https://www.google.com/maps/@53.2124089,-2.73569,3a,75y,130.94h/data=!3m4!1e1!3m2!1sAF1QipOjLYuWo-mQx1ZQgQ2NxInDZcKunJNFsNSdBGU!2e10 Check out England, GBR Shared via the #StreetView app

High mounted number plates on the back of tractors.

https://www.google.com/maps/@53.2100342,-2.7415692,3a,75y,285.6h/data=!3m4!1e1!3m2!1sAF1QipNbYaDVqc0cnnve1gXel_NpkubxQzhpPGMj7eo!2e10 Check out England, GBR Shared via the #StreetView app

In your solution please consider both style and colour of number plate and location on the vehicle.

This particular capture was also a serious work out for the face blurring algorithm. On the whole it only missed a few faces.

HQ | Blog | Spotted a Trekker? | Become a Trekker | Facebook | Instagram