Image processing · Digital imaging

Has anybody used/come across image recognition solutions?

Sujit Hemadri Nobody

December 23rd, 2015

Looking basically to obtain the dimensions of the object captured from phone camera/external device. Would be of help if somebody can throw some light on this.

Eyal Feingersch

December 23rd, 2015

Unless there is a reliable algorithm that can estimate distance from a 'normal' camera, you need to employ a combination that includes a depth camera.
Just like Kineckt and Intel Realsense, in order to calculate the Z axis.

Most devices don't have such camera, even none at all.

My work includes development on Realsense, and I AFAIK, the Creative R200 camera, which is the 'world facing' 3d camera that is supposed to be integrated into mobile devices, is not available yet.

Steve Donie Principal Software Engineer at Datical

December 23rd, 2015

Maybe see if you can talk to this guy: 

http://www.bloomberg.com/features/2015-george-hotz-self-driving-car/

Massimiliano Versace CEO & President, Neurala Inc.; Director, Boston University Neuromorphics Lab

December 23rd, 2015

Yes, we specialize on solutions that use brain-inspired algorithms to enable perception/navigation in robotic platforms and other devices using off-the-shelf cameras and IMUs. www.neurala.com
 The devil is in the details though... Happy to answer specific questions after I learn ab bit more about the specific use case and sensors you are using. Versace@neurala.com

Mark Hill Founder & CEO at The Wardrobe Essentialist

December 23rd, 2015

If you can influence the photo, including a common reference object of known size (a coin, Dvd, etc) makes the problem much easier. 

Sujit Hemadri Nobody

December 23rd, 2015

Thanks Mark for that reply! Have been thinking on those lines. It can be done for static objects and those whose sizes increases in fixed & known ways. Becomes difficult where sizes are varying in multiple ways. Eg human body dimensions.

Joanan Hernandez CEO & Founder at Mollejuo

December 23rd, 2015

Hello,

Google's project Tango does something along the line.

Good luck!

Cheers!

Alexx Henry CEO BlueVishnu, Avatar Shepherd

December 26th, 2015

You can use photogrammetry to obtain depth and object dimensions using any RGB camera without the need for depth sensors. The accuracy will be dependent on a number of factors and is much more limited if you are using a singe camera (especially when there's movement on the camera and source). We use dozens of fixed cameras for human subjects but we're interested in sub mm accuracy without optical flow. 

 A colleague, Michael Hollroyd and his team at Arqball have been using single camera photogrammetry techniques using camera phones. Maybe worth investigating. https://www.linkedin.com/company/1437849?trk=prof-0-ovw-curr_pos

Deanan DaSilva Technology and Product Development Consultant

December 27th, 2015

Have a look at openCV for mobile.

https://www.youtube.com/watch?v=bcswZLwhTUI

Cheryl Greene Co-founder and Executive Producer at DrGreene.com

January 29th, 2016

If you're still looking, I think a Boston machine learning start-up might have a solution --> http://www.indico.com/

Good luck..