Skip to content
  • The Student Engineer
  • C2I Awards
  • Salary Survey
  • Latest Issue
  • Newsletter Signup
  • Employer Zone
  • Covid-19
The Engineer
  • News
  • In-depth
  • Opinion
  • JOBS
  • Sectors
  • Supplier Network
Search Login / Register Primary Menu
Close
Login or Register
News Automotive Electronics & communications Robotics

AI program coverts 2D into 3D

26th July 2017 11:09 am 26th July 2017 11:24 am

Researchers at Purdue University have developed a new program that uses artificial intelligence to transform 2D images into 3D shapes.

Known as SurfNet, the program could have applications in robotics and autonomous vehicles, as well as content creation for the emerging 3D industry. It works by mapping 2D images onto a 3D sphere, then opening up that sphere to create a new 2D shape.

SurfNet then applies that technique to huge numbers of 2D images, learning abstractly via machine and deep learning. The system learns the 3D image and the 2D image in pairs, and this enables it to predict other, similar 3D shapes from just a 2D image.

“If you show it hundreds of thousands of shapes of something such as a car, if you then show it a 2D image of a car, it can reconstruct that model in 3D,” said Karthik Ramani, Purdue’s Donald W Feddersen Professor of Mechanical Engineering

“It can even take two 2D images and create a 3D shape between the two, which we call ‘hallucination.'”

According to Ramani, the technique allows for greater accuracy and precision than current 3D deep learning methods that use volumetric pixels, or voxels.

“We use the surfaces instead since it fully defines the shape,” he explained. “It’s kind of an interesting offshoot of this method. Because we are working in the 2D domain to reconstruct the 3D structure, instead of doing 1,000 data points like you would otherwise with other emerging methods, we can do 10,000 points. We are more efficient and compact.”

“This is very similar to how a camera or scanner uses just three colours, red, green and blue—known as RGB—to create a colour image, except we use the XYZ coordinates.”

As the program improves, the researchers hope it can be used to create 3D content using non-specialist 2D equipment, and also allow machines to understand 3D environments from simple 2D images.

The Engineer Jobs

Visit the UK’s dedicated jobsite for engineering professionals. Each month, we’ll bring you hundreds of the latest roles from across the industry.

View jobs
Opinion Skills and careers

Engineering: graduate salaries send a worrying message to the top students

1st January 2015 12:00 am 20th February 2018 11:50 am
News Robotics

Video of the week: Stanford tubular robot grows like a vine

20th July 2017 10:59 am 20th July 2017 3:26 pm
Blog Rail & marine

Is Hyperloop more than just loopy hype?

14th July 2017 1:24 pm 14th July 2017 1:47 pm
Latest ArticlesComments (1)
News Renewables

Assessing renewable energy sites with drones

5th March 2021 11:52 am 5th March 2021 3:31 pm
offshore wind farm
News Energy & environment

Sea mussels to inspire offshore wind farm solutions

5th March 2021 11:25 am 5th March 2021 11:25 am
Opinion Policy & business

Can ARIA save UK innovation? asks RAE President Sir Jim McDonald

5th March 2021 8:30 am 5th March 2021 2:19 pm
niche
Opinion Manufacturing

Promoted content: Niche new product introduction 

5th March 2021 7:00 am 5th March 2021 12:02 pm
Comments
  • John Rupkalvis 27th July 2017 at 8:52 pm

    This system requires the program to anticipate based on prior knowledge. Thus, it may prove to be a very useful tool in the conversion system, but it cannot be applied to everything. When the requirement is only to produce a “believable” result, it may work well for many objects, but even there, likely the differences from one object to another may often require some human intervention or “tweaking” with some objects. This may be most useful for “entertainment” applications. If, however, the requirement is for an accurate metrological result that can actually be accurately measured in all three axes, such as for medical, industrial, and scientific applications, then this approach may be less useful. Rather than replacing existing methodology, I see this as another, possibly useful, tool in the conversion toolbox.

    Reply Link
  • Post a comment
    Cancel reply

    Threaded commenting powered by interconnect/it code.

    All comments are moderated. Click here for our guidelines.

Explore

  • SECTORS
    • AEROSPACE
    • AUTOMOTIVE
    • CIVIL AND STRUCTURAL
    • DEFENCE AND SECURITY
    • ELECTRONICS AND COMMUNICATIONS
    • ADVANCED MANUFACTURING
    • MATERIALS
    • MEDICAL AND HEALTHCARE
    • RAIL AND MARINE
    • ENERGY AND ENVIRONMENT
      • NUCLEAR
      • COAL, OIL AND GAS
      • ENERGY AND ENVIRONMENT
      • RENEWABLES
  • CAREERS
    • THE STUDENT ENGINEER
    • JOBS
    • SALARY CALCULATOR
    • SALARY SURVEY 2018
  • SKILLS AND CAREERS
  • POLICY & BUSINESS
  • VIDEOS
  • EVENTS
  • WEBINARS
  • CLASSIC ARCHIVE
  • C2I 2020 Winners Book
  • BACK ISSUES
  • NEWSLETTER SIGNUP
  • TECH TRENDS 2021
  • Twitter
  • LinkedIn
  • Instagram
  • Flip
  • Advertise
  • Contact us
  • Cookies
  • Digital edition
  • Magazine subscription
  • Privacy
  • Website Terms and Conditions

Copyright © Mark Allen Engineering Ltd (a Mark Allen Group company) 2019

Mark Allen Engineering Limited
Registered Office: Mark Allen Group, St Jude's Church, Dulwich Road, London, SE24 0PB
Registered in England No. 11569365