Face Recognition

[Viktor_B_]

Hello,

I want to create a solution to identify a person by a face image. This should work with regular webcams. And it should not be just a toy.

I'm not an expert in this area and I have a question. Is it possible to do this using Intel RealSense SDK 2014?

For example, the previous version supported only 7 landmark points, I think that it was insufficient. The current version supports 78 landmark points. It gives a hope. But is that enough?

[Colleen_C_Intel]

The answer to the question of "what is enough for facial recognition?" depends on the circumstances of your use models and your security demands. 

Telling users apart in a game is much less demanding (security wise) than allowing access to remotely manage all systems on a corporate network. But the Real Sense software has the many landmarks and using a depth camera, it can include more data than 2D cams, making it much less likely to be fooled by a photograph or mask. 

For discussion on security, I suggest you try security forums.

[Pubudu-Silva_Intel]

Colleen is right, it all depends on what you want to do and how you want to do it. Just to add to her answer,

If your face authentication requirement is rather lax, then you can be creative and do something with landmarks given to you  buy the SDK. If you need more precision consider using eyes for authentication, and there are ample of research publications on how to do that. In this case you have to right your own code for eye authentication, but the SDK can help you in locating eye region so you can run your algorithms on a selected pixels making it less computationally intensive. 

If you consider using a depth camera, you have the luxury of using facial depth information too. The cool thing about the SDK architecture is that it is designed to provide you necessary general tools to build an specific application like this with lot of freedom to innovate 

[Viktor_B_]

Colleen Culbertson, PUBUDU

Thanks for replied. I need to think about your comments.