I'll take your frame to another dimension, pay close attention: This AI auto-generates 3D objects from 2D snaps

Drivers, translators, factory workers, programmers, IT pros... and now animators


Boring 2D images can be transformed into corresponding 3D models and back into 2D again automatically by machine-learning-based software, boffins have demonstrated.

The code is known as a differentiable interpolation-based renderer (DIB-R), and was built by a group of eggheads led by Nvidia. It uses a trained neural network to take a flat image of an object as inputs, work out how it is shaped, colored and lit in 3D, and outputs a 2D rendering of that model.

This research could be useful in future for teaching robots and other computer systems how to work out how stuff is shaped and lit in real life from 2D still pictures or video frames, and how things appear to change depending on your view and lighting. That means future AI could perform better, particularly in terms of depth perception, in scenarios in which the lighting and positioning of things is wildly different from what's expected.

Jun Gao, a graduate student at the University of Toronto in Canada and a part-time researcher at Nvidia, said: “This is essentially the first time ever that you can take just about any 2D image and predict relevant 3D properties.”

During inference, the pixels in each studied photograph are separated into two groups: foreground and background. The rough shape of the object is discerned from the foreground pixels to create a mesh of vertices.

Next, a trained convolutional neural network (CNN) predicts the 3D position and lighting of each vertex in the mesh to form a 3D object model. This model is then rendered as a full-color 2D image using a suitable shader. This allows the boffins to compare the original 2D object to the rendered 2D object to see how well the neural network understood the lighting and shape of the thing.

AI_lego_sorter

You looking for an AI project? You love Lego? Look no further than this Reg reader's machine-learning Lego sorter

READ MORE

During the training process, the CNN was shown stuff in 13 categories in the ShapeNet dataset. Each 3D model was rendered as 2D images viewed from 24 different angles to create a set of training images: these images were used to show the network how 2D images relate to 3D models.

Crucially, the CNN was schooled using an adversarial framework, in which the DIB-R outputs were passed through a discriminator network for analysis.

If a rendered object was similar enough to an input object, then DIB-R’s output passed the discriminator. If not, the output was rejected and the CNN had to generate ever more similar versions until it was accepted by the discriminator. Over time, the CNN learned to output realistic renderings. Further training is required to generate shapes outside of the training data, we note.

As we mentioned above, DIB-R could help robots better detect their environments, Nvidia's Lauren Finkle said: “For an autonomous robot to interact safely and efficiently with its environment, it must be able to sense and understand its surroundings. DIB-R could potentially improve those depth perception capabilities.”

The research will be presented at the Conference on Neural Information Processing Systems in Vancouver, Canada, this week. ®

Similar topics


Other stories you might like

  • What's the big deal with service meshes? Think of them as SDN at Layer 7

    A technical yet demystifying dive into networking tech you now can't avoid

    Systems Approach I remember when I first heard about Service Meshes in 2017, and wondering what the big deal was. Building cloud applications as a graph of microservices was commonplace, and telcos were hard at work inventing yet other ways to chain together virtualized network functions. Service graphs, service chains, service meshes… how many ways do we really need to talk about composing complex systems from a collection of smaller components?

    It wasn’t until I recognized a familiar pattern that I got it: a Service Mesh is just SDN at Layer 7. That’s probably what happens when SDN is the hammer you keep hitting nails with, but I’ve come to believe there is value in that perspective.

    The figure below highlights the similarities between the two scenarios, both of which include a centralized controller that issues directives to a distributed set of connectors (physical/virtual switches in one case, and a sidecar container in the other case) — based on a combination of policy intents from above and monitoring data reported from below. The primary difference is that the SDN controller on the left is controlling L2/3 connectivity and the Service Mesh on the right is controlling L7 connectivity.

    Continue reading
  • Mars race: China dreams of nuclear rockets, manned bases, and space elevators

    We're looking forward to the late 21st-century colony wars

    Over the next quarter century, China wants to set up a permanent base on Mars for "large scale development of the Red Planet," and install a sci-fi carbon-nanotube elevator to shuttle goods between the surface and spacecraft in orbit.

    That’s according to the China Academy of Launch Vehicle Technology (CALT), the country’s largest rocket maker, which described a road-map outlining the Middle Kingdom's ambition to explore the unforgiving dust world. Missions to Mars are planned for 2033, 2035, 2037, 2041, and 2043 quite possibly using nuclear-propelled spacecraft.

    In a speech, CALT’s President Wang Xiaojun said his state-owned organization first intends to send robots to Mars to collect samples of material to study back on Earth. These machines will also scout out good locations to develop a human settlement.

    Continue reading
  • Bridging the observability gap

    Trace the journey through all those microservices in the background

    Sponsored In modern IT, visibility is everything. IT admins and Site Reliability Engineers (SRE) survive on their ability to see what's happening in their systems. Unfortunately, as systems get more sophisticated, it has become harder to see what they're doing. That's why the industry is promoting observability as the evolution of existing concepts like monitoring and metrics. Vendors are stepping up with tools to address a growing visibility gap.

    Continue reading
  • Google: About that whole getting rid of third-party cookies thing – we're gonna need another year or so

    Plan to reinvent advertising turns out to be more difficult than expected

    Google, which makes the only major browser not blocking third-party cookies by default, has revised its commitment to phase out third-party cookies by 2022.

    The super-corp's biscotticide is now scheduled to begin in mid-2023 and run through late 2023.

    Third-party cookies refer to tracking files deposited in one's browser when visiting a website that includes code interacting with third-party domains. The firms associated with these domains, typically marketing and analytics businesses, check for the presence of their cookies across different websites and use this information to build marketing profiles and to target ads based on behavior.

    Continue reading
  • These six proposed bipartisan antitrust laws put Big Tech in the cross-hairs – and a House committee just OK'd them

    Well, it's a start

    The US House Judiciary Committee this week approved half a dozen major bipartisan antitrust bills aimed at clamping down on the growing power of Big Tech and its monopolization of some markets.

    The panel, led by Jerry Nadler (D-NY), debated for nearly 30 hours on Wednesday and Thursday to advance the wide-sweeping six-bill package. The proposed laws includes all sorts of measures to prevent companies like Google, Apple, Amazon, Microsoft, Facebook, and others from dominating their sectors of the technology industry.

    There was likely plenty of lobbying and other wrangling going on in the back and foreground over the exact wording of the package. For instance, there was a concern by some lawmakers that Microsoft would end up avoiding certain provisions in the proposed acts that would otherwise hit Google and Apple. Tweaks were made – such as removing "mobile" from "mobile operating system" in the fine-print – to ensure no one was wriggling out.

    Continue reading
  • You won't want that Linux bling if it comes from Pling: Marketplace platform has critical vulnerabilities

    No one wants to be pwned by a drive-by RCE

    A Berlin startup has disclosed a remote-code-execution (RCE) vulnerability and a wormable cross-site-scripting (XSS) flaw in Pling, which is used by various Linux desktop theme marketplaces.

    Positive Security, which found the holes and is not to be confused with Russia’s Positive Technologies, said the bugs are still present in the Pling code and its maintainers have not responded to vulnerability reports.

    Pling presents itself as a marketplace for creative folk to upload Linux desktop themes and graphics, among other things, in the hope of making a few quid from supporters. It comes in two parts: code needed to run your own bling bazaar, and an Electron-based app users can install to manage their themes from a Pling souk. The web code has the XSS in it, and the client has the XSS and an RCE. Pling powers a bunch of sites, from pling.com and store.kde.org to gnome-look.org and xfce-look.org.

    Continue reading
  • Would-be password-killer FIDO Alliance aims to boost uptake with new UX guidelines

    Throws a bone to complex enterprise deployment, too

    The FIDO Alliance, which operates with no smaller mission than to "reduce the world's over-reliance on passwords", has announced the release of new user experience (UX) guidelines aimed at bringing the more technophobic on board.

    Launched back in 2013 as the Fast Identity Online Alliance, the FIDO Alliance aims to do away with passwords altogether through the introduction of standards-compliant "authenticators" including USB security dongles, fingerprint readers, Trusted Platform Modules (TPMs) and more.

    While the organisation's standards, which were updated with the launch of FIDO2 in 2018, have enjoyed adoption in the majority of web browsers and with a range of companies, they're still seen as unusual and even inconvenient compared to the good ol' username and password combo – which is where the new UX guidelines come in.

    Continue reading
  • UK's Vodafone network runs trials on standalone 5G in London, Manchester and Cardiff

    These are networks that are not dragged down by LTE core

    Vodafone has launched 5G SA (Standalone) trials in London, Manchester, and Cardiff in its largest test of the technology yet.

    The commercial launch has allowed the carrier to experiment with new ways to commercialise its network, including network slicing – where a portion of network is dedicated to a specific customer for their exclusive use. It will also allow customers to test 5G SA devices on a live, public network.

    Vodafone selected Ericsson's dual-mode 5G core network as the dedicated provider for this trial. It follows trials at Coventry University in 2020, and a separate trial in Spain.

    Continue reading
  • What you need to know about Microsoft Windows 11: It will run Android apps

    The operating system they said shouldn't exist

    Microsoft on Thursday announced Windows 11, or tried to as an uncooperative video stream left many viewers of the virtual event flummoxed by intermittent transmission gaps in the opening minutes.

    The technical issues proved bad enough that Matt Velloso, Technical Advisor to the CEO at Microsoft, suggested trying the YouTube video stream as an alternative to the Microsoft-hosted one.

    But with some of the features already known as a result of a leaked build last week, the impact of the intermittent video dropouts was less than it might have been.

    Continue reading
  • Russia spoofed AIS data to fake British warship's course days before Crimea guns showdown

    Great powers clash while the rest of us sigh and tut at data feed meddling

    Russia was back up to its age-old spoofing of GPS tracks earlier this week before a showdown between British destroyer HMS Defender and coastguard ships near occupied Crimea in the Black Sea.

    Yesterday Defender briefly sailed through Ukrainian waters, triggering the Russian Navy and coastguard into sending patrol boats and anti-shipping aircraft to buzz the British warship in a fruitless effort to divert her away from occupied Crimea's waters.

    Russia invaded Ukraine in 2014 and has occupied parts of the region, mostly in the Crimean peninsula, ever since. The UK and other NATO allies do not recognise Ukraine as enemy-held territory so Defender was sailing through an ally's waters – and doing so through a published traffic separation scheme (similar to the TSS in the English Channel), as Defence Secretary Ben Wallace confirmed this afternoon.*

    Continue reading
  • Lego bricks, upcycled iPhone lenses used in new low-cost, high-res microscope

    Full instructions given away for free, to 'nurture natural curiosity'

    A trio of boffins at the Georg August University Göttingen and Münster University have put together a low-cost yet high-resolution microscope for educational users – using smartphone parts and Lego bricks.

    "An understanding of science is crucial for decision-making and brings many benefits in everyday life, such as problem-solving and creativity," said Timo Betz, professor at the University of Göttingen and co-author of the paper detailing the project. “Yet we find that many people, even politicians, feel excluded or do not have the opportunities to engage in scientific or critical thinking.

    "We wanted to find a way to nurture natural curiosity, help people grasp fundamental principles and see the potential of science."

    Continue reading

Biting the hand that feeds IT © 1998–2021