Return to Kuva ja ääni

Kuva- ja äänilinkkejä

CNET esittelyt


  • Researchers teach an AI to generate logical images based on text captions

    The Allen Institute for AI (AI2) created by Paul Allen, best known as co-founder of Microsoft, has published new research on a type of artificial intelligence that is able to generate basic (though obviously nonsensical) images based on a concept presented to the machine as a caption. The technology hints at an evolution in machine learning that may pave the way for smarter, more capable AI.

    The research institute's newly published study, which was recent highlighted by MIT, builds upon the technology demonstrated by OpenAI with its GPT-3 system. With GPT-3, the machine learning algorithm was trained using vast amounts of text-based data, something that itself builds upon the masking technique introduced by Google's BERT.

    Put simply, BERT's masking technique trains machine learning algorithms by presenting natural language sentences that have a word missing, thus requiring the machine to replace the word. Training the AI in this way teaches it to recognize language patterns and word usage, the result being a machine that can fairly effectively understand natural language and interpret its meaning.

    Building upon this, the training evolved to include an image with a caption that has a missing word, such as an image of an animal with a caption describing the animal and the environment -- only the word for the animal was missing, forcing the AI to figure out the right answer based on the sentence and related image. This taught the machine to recognize the patterns in how visual content related to the words in the captions.

    This is where the AI2 research comes in, with the study posing the question: 'Do vision-and-language BERT models know how to paint?'

    Experts with the research institute build upon the visual-text technique described above to teach AI how to generate images based on its understanding of text captions. To make this possible, the researchers introduced a twist on the masking technique, this time masking certain parts of images paired with captions to train a model called X-LXMERT, an extension of the LXMERT model family that uses multiple encoders to learn connections between language and visual data.

    The researchers explain in the study [PDF]:

    Interestingly, our analysis leads us to the conclusion that LXMERT in its current form does not possess the ability to paint - it produces images that have little resemblance to natural images ...

    We introduce X-LXMERT that builds upon LXMERT and enables it to effectively perform discriminative as well as generative tasks ... When coupled with our proposed image generator, X-LXMERT is able to generate rich imagery that is semantically consistent with the input captions. Importantly, X-LXMERT’s image generation capabilities rival state-of-the-art image generation models (designed only for generation), while its question-answering capabilities show little degradation compared to LXMERT.

    By adding the visual masking technique, the machine had to learn to predict what parts of the images were masked based on the captions, slowly teaching the machine to understand the logical and conceptual framework of the visual world in addition to connecting visual data with language. For example, a clock tower located in a town is likely surrounded by smaller buildings, something a human can infer based on the text description.

    An AI-generated image based on the caption, 'A large painted clock tower in the middle of town.'

    Using this visual masking technique, the AI2 researchers were able to impart the same general understanding to a machine given the caption, 'A large clock tower in the middle of a town.' Though the resulting image (above) isn't realistic and wouldn't be mistaken for an actual photo, it does demonstrate the machine's general understanding of the meaning of the phrase and the type of elements that may be found in a real-world clocktower setting.

    The images demonstrate the machine's ability to understand both the visual world and written text and to make logical assumptions based on the limited data provided. This mirrors the way a human understands the world and written text describing it.

    For example, a human, when given a caption, could sketch a concept drawing that presents a logical interpretation of how the captioned scene may look in the real world, such as computer monitors likely sitting on a desk, a skier likely being on snow and bicycles likely being located on pavement.

    This development in AI research represents a type of simple, child-like abstract thinking that hints at a future in which machines may be capable of far more sophisticated understandings of the world and, perhaps, any other concepts they are trained to understand as related to each other. The next step in this evolution is likely an improved ability to generate images, resulting in more realistic content.

    Using artificial intelligence to generate photo-realistic images is already a thing, though generating highly specific photo-realistic images based on a text description is, as shown above, still a work in progress. Machine learning technology has also been used to demonstrate other potential applications for AI, such as a study Google published last month that demonstrates using crowdsourced 2D images to generate high-quality 3D models of popular structures.

  • Slideshow: Winners of the 2020 Drone Photo Awards from Siena Awards

    Winners of the 2020 Drone Photo Awards from Siena Awards

    Winners for the 6th edition of the Drone Photo Awards competition, affiliated with the Siena International Photo Awards competition (you can view winners of the Creative Photo Awards here), were recently announced. Entries were sent in by drone photographers from 126 countries. 'Love Heart of Nature' by Australian photographer Jim Picôt, which depicts a shark swimming inside a heart–shaped salmon school, was recognized as the Overall Winner.

    The awards are divided into 9 categories: Abstract, Empty Cities: Life during COVID-19, Nature, People, Sports, Series, Urban Architecture, Wedding, and Animals. All 45 winning images will be displayed at the 'Above Us Only Sky' exhibition, scheduled from October 24th to November 29th at the 'Accademia dei Fisiocritici' museum in Siena, Italy.

    Overall Winner: 'Love Heart of Nature' by Jim Picôt

    Location: Avoca Beach, NSW, Australia

    Description: In winter, a shark is inside a salmon school when, chasing the baitfish, the shape became a heart shape.

    Winner, Wedding: 'Tropical Bride' by Mohamed Azmeel

    Location: (Not given)

    Description: I used the flowers and the leaves leftover from the decoration of a wedding, to make something creative.

    Winner, Abstract: 'Swirl' by Boyan Orste

    Location: Pink Lake, Australia

    Description: An abstract shot of a Pink lake chemical reaction in Western Australia.

    Winner, Nature: 'Coffee or Tea' by Yi Sun

    Location: Brazil

    Description: (Not given)

    Winner, Wildlife: 'Outer Space Flamingos' by Paul McKenzie

    Location: Lake Natron, Tanzania

    Description: (Not given)

    Winner, Life Under COVID-19: 'Black Flag' by Tomer Appelbaum

    Location: Israel

    Description: Thousands of Israelis maintain social distancing due to Covid-19 restrictions while protesting against Israeli Prime Minister Benjamin Netanyahu in Rabin Square on 19 April 2020.

    Winner, Sport: 'On the Sea' by Roberto Corinaldesi

    Location: Cornwall, United Kingdom

    Description: An aerial view of swimmers, where the sea becomes the place to take refuge, between the blue carpet and the white foam of the waves.

    Winner, People: 'Frozen Land' by Alessandra Meniconzi

    Location: Eurasian Steppe

    Description: With temperatures of minus 30°C, winters in the Eurasian steppe can be brutal. But life doesn't stop, and local people move from one village to another with a sledge, crossing icy rivers and lakes.

    Winner, Urban: 'Alien Structure on Earth' by Tomasz Kowalski

    Location: Kuala Lumpur, Malaysia

    Description: Sometimes we need to change the perspective to feel the strength of the structure stronger than we've ever thought. The Petronas Towers, also known as the Petronas Twin Towers, are twin skyscrapers in Kuala Lumpur.

    Winner, Wedding: 'The Wedding Crashers' by David Gallardo

    Location: Turks & Caicos Islands

    Description: (Not given)

    Winner, Life Under COVID-19: 'Lonely Guardian' by Mauro Pagliai

    Location: Siena, Italy

    Description: (Not given)

    Winner, Sport: 'Ball Up' by Brad Walls

    Location: Sydney, Australia

    Description: The physical motions of the tennis player against the clean abstract lineage of the court created a harmonious effect to the eye.

    Winner, Abstract: 'Fishing At Jamuna River' by MD Tanveer Hassan Rohan

    Location: Bogra, Bangladesh

    Description: (Not given)

    Winner, Urban: 'Sunrise on the Top' by Rex Zou

    Location: Shanghai, China

    Description: At 4:30 in the morning, mysteriously shrouded in clouds, this is what the second tallest building in Shanghai looks like.

    Winner, People: 'Mountains of Salt' by Igor Altuna

    Location: Thi Xa Ninh Hoa, Vietnam

    Description: An aerial picture taken on a saltern near a small town on central Vietnam’s coast.

  • Nikon Z6 II and Z7 II coming October 14

    Nikon has announced, via a countdown on its website, it will be launching its next-generation full-frame mirrorless cameras on October 14.

    Absolutely no information is given about the Z7 II and Z6 II cameras aside from their respective names, which eschew Nikon’s usual ’S’ naming scheme for the more Canon-esque ‘Mark II’ versions.

    We’ll be sure to keep you up to date on the latest developments from Nikon.

  • Canon purchases new supercomputer system to further its 'no-prototype' product development ambitions

    Fujitsu Limited announced last week that Canon Inc. has ordered a new supercomputer. The purchase signals that Canon is progressing with its plans to develop products without the need for prototypes, thereby reducing development costs and increasing development efficiency.

    The system Canon ordered includes a Fujitsu Supercomputer PRIMEHPC FX1000 unit. Canon has previously used Fujitsu PRIMEHPC FX10 and PRIMEHPC FX100 supercomputer systems. The new FX1000 unit includes class-leading technology from Fugaku, the world's fastest supercomputer. Fugaku, a petascale supercomputer, was jointly developed by RIKEN and Fujitsu starting in 2014. Fugaku is scheduled to begin operation next year. When tested in June, it became the fastest supercomputer in the world. Fugaku has an expected theoretical computational performance of 648.8 teraflops.

    Fujitsu Limited also announced that upon its completion, the supercomputer headed to Canon will play a critical role in Canon's 'no-prototype' product development plans. The supercomputer will allow for 'enhanced capabilities and scope of applicability of analysis in Canon's product development process.' Fujitsu and Canon anticipate the supercomputer to begin operations during the first half of 2021.

    Sample image of a simulation created using an existing supercomputer. This is a simulation of a Canon inkjet printer being dropped while in its packaging. With the new FX1000, Canon will be able to perform much more complicated simulations of impacts and much more. Image credit: Canon.

    Canon's 'no-prototype' initiative applies to the development of numerous products, including cameras. Canon is utilizing 3D CAD data in 'analytical simulations to evaluate multiple facets of proposed products, including functionality, as well as ease of manufacturing.' The PRIMEHPC FX1000 system, which includes 192 nodes, will allow Canon to produce larger scale analyses and simulations, including simulations with over 100 million total elements. The new system will also Canon to produce analyses of simulated airflow and electromagnetic waves.

    In its coverage of this announcement, PetaPixel speculates that Canon is working hard to catch up to the fast and furious development cycle of Sony and its mirrorless cameras and that a new supercomputer can speed up Canon's development cycles.

    Fujitsu PRIMEHPC supercomputer have numerous applications, including aiding in establishing a product development and manufacturing cycle that isn't reliant upon prototypes. Image credit: Fujitsu

    It stands to reason that if Canon can quickly simulate many iterations of a product it is developing rather than need to construct real-world prototypes, the revision process will be quicker. Further, leveraging a class-leading supercomputer may allow for deep analyses of products that may have otherwise taken Canon far too long or been outright impossible.

    Ultimately, how the FX1000 supercomputer system will impact Canon's product development cycles remains to be seen. The system should be operational in the first half of 2021, but of course, the development cycle of something like a new camera is quite lengthy.

  • Atomos launches Connect, a $79 capture card

    Atomos has launched a HDMI-to-USB video capture card that allows any HDMI camera or recorder to be used to live stream or record to a PC or Mac. Connect is a tiny device that plugs directly into a USB port and accepts HDMI input to provide a video feed of up to 1080p at 60fps in up to 12-bit color.

    The device can accept 4K at 30p, and needs no additional drivers, software or power supply to work. And best of all, it costs $79. Unlike most capture cards, the Connect doesn’t use a USB out cable, so the full weight of the HDMI cable will be borne by the USB port on your computer, making this perhaps a solution when a short cable is in use or when the cable can be supported.

    Atomos is promoting Connect to use with its range of video recorders, but it can also work directly with any camera, or device, that can output via HDMI. The company points out though that when used with the Shogun7 users can switch between up to four simultaneously streaming cameras using the Shogun’s multiple-camera recording capabilities.

    The Atomos Connect is available now. For more information see the Atomos website.

    Press release:

    Atomos launch Connect: Professional HDMI to USB conversion for streaming.

    Melbourne, Australia – 28th September 2020 – Atomos is today excited to launch Connect, a convenient, reliable, and affordable bridge between professional 4K video capture and high frame rate UVC streaming over USB (up to 1080p60). Connect works with any HDMI device including the entire Atomos range. Simple to use, it has universal support from popular streaming platforms and at only $79USD is an affordable way for gamers, vloggers and education facilities to up their streaming game using the professional Atomos monitor-recorder ecosystem.

    Convert, Connect, Stream
    Connect bridges the Atomos world of professional recording and monitoring with popular streaming platforms such as YouTube, Twitch and OBS. Use your Atomos device for professional shot setup and recording and use Connect to simply convert the HDMI output to USB on the go for streaming on your Mac or PC. Connect accepts up to 4Kp30 video and is capable of up to 1080p60 output over USB – incredible streaming quality for the US$79 price point.

    Plug and go like a pro
    With a slim 0.5-inch profile, Connect is sleek enough to sit directly off your laptop or computer. No drivers are required and no power supply either, meaning it really is a simple matter of connecting your Atomos or HDMI device to your Mac or PC to start producing like a pro.

    Shogun7 Connect – tame multicamera streaming
    The Atomos Shogun7 allows you to monitor and record up to four 1080p60 streams simultaneously and then cue and switch in real-time with a simple touch of the screen. Additionally, the Shogun7 is able to accept asynchronous feeds (sources don’t need to be locked), allowing the user to use more affordable cameras for multi camera production. Adding in Connect, lets you output the switched feed for live streaming on your Mac or PC. This makes the Shogun7 Connect package one of the easiest and most affordable options for professional switching.

    Transforms your Camera to a Pro Webcam
    Even without an Atomos monitor-recorder, Connect transforms your HDMI camera to a powerful professional webcam. No drivers required, plug in Connect and go live instantly.

    The Atomos Connect is available immediately for $79/€79 from the global authorised Atomos dealer network.

  • Video: Adobe teases Advance Color Grading tool coming to ACR, Lightroom and Lightroom Classic

    Adobe has announced a new Advanced Color Grading feature will soon be available inside Adobe Camera Raw, Lightroom and Lightroom Classic.

    The new Color Grading panel is inspired by the Lumetri Color panel used in Adobe’s Premiere Pro video editing software and replaces the previous Split Toning panel. Unlike Split Toning, which only allows you to adjust the highlights and shadows, the new Color Grading panel takes it one step further, allowing you to also adjust the coloring of the midtones.

    In addition to adding midtones, the panel is also redesigned. It now uses a three-wheel system, not unlike the color wheels often seen in professional video editing programs. Each wheel works alongside a slider to provide complete HSL control for highlights, midtones and shadows. There’s also a global color wheel for more general adjustments, as well as a blending slider to help boost or reduce your adjustments, depending on the look you’re going for.

    Adobe doesn’t mention when we can expect to see this filter in Adobe Camera Raw, Adobe Lightroom and Adobe Lightroom Classic, but does say it’ll be showing off more of this new panel at this year’s virtual Adobe Max conference, which is taking place on October 20–22.

  • Venus Optics brings its Laowa 9mm T2.9, 12mm T2.9 and 15mm T2.1 cine primes to Canon RF mount

    Venus Optics has announced it’s now adding a Canon RF lens mount option for three of its ultra-wide cinema primes: the 9mm T2.9 ‘Zero-D’ lens, 12mm T2.9 ‘Zero-D’ lens and 15mm T2.1 ‘Zero-D.’

    The 12mm T2.9 ‘Zero-D’ lens was released back in January of this year, while the 9mm T2.9 and 15mm T2.1 ‘Zero-D’ lenses were announced back in June of this year. At the time, the lenses were only available for Sony E, Canon EF and PL mount, but owners of Canon’s EOS R series mirrorless cameras can now get native versions of the lenses, negating the need to use the EF version with the EF-RF adapter.

    Photo Credit: Photo by Phil Holland, provided by Venus Optics

    As part of the Loawa Zero-D’ line, all three lenses offer ‘close-to-zero’ distortion. All three lenses also feature industry standard 0.8 mod pitch gears for controlling both the aperture and focus.

    As a quick refresher, below are specification summaries for each of the lenses:

    9mm T2.9 ‘Zero-D’ lens

    The 9mm T2.9 is constructed of 15 elements in 10 groups, has a minimum focusing distance of 12cm (4.72”), uses a seven-blade aperture diaphragm and weighs 247g (8.71oz).

    12mm T2.9 ‘Zero-D’ lens

    The 12mm T2.9 is constructed of 16 elements in 10 groups, has a minimum focusing distance of 18cm (7.09”), uses a seven-blade aperture diaphragm and weighs 675g (1.5lbs).

    15mm T2.1 ‘Zero-D’ lens

    The 15mm T2.1 is constructed of 12 elements in 9 groups, has a minimum focusing distance of 15cm (5.91”), uses a seven-blade aperture diaphragm and weighs 540g (1.19lbs).

    All three lenses are available to purchase with the Canon RF mount starting today on Laowa’s online shop. The 9mm T2.9, 12mm T2.9 and 15mm T2.1 retail for $600, $1,500 and $1,200, respectively.

  • Home on the range: Exploring Sony's AF technologies on people and animals

    Sony's current range of Alpha and RX-series cameras are packed with the company's latest autofocus technologies. These include highly advanced face and eye-detection for both human and animal subjects.

    Photographer and Rancher Alyssa Henry lives with her family on her ranch in Montana. A perfect location – and perfect subject-matter – to put Sony's latest autofocus technologies to the test.

    Home on the range: Sample images $(document).ready(function() { SampleGalleryV2({"containerId":"embeddedSampleGallery_8639258908","galleryId":"8639258908","isEmbeddedWidget":true,"selectedImageIndex":0,"isMobile":false}) });

    This is sponsored content, created with the support of Amazon and Sony. What does this mean?

  • Sample gallery: macro photos with a motorized slider

    $(document).ready(function() { SampleGalleryV2({"containerId":"embeddedSampleGallery_2283835564","galleryId":"2283835564","isEmbeddedWidget":true,"selectedImageIndex":0,"isMobile":false}) });

    Macro photography has captured the imagination of many-a photographer looking to get out of a creative slump. It's an eye-opening experience to see otherwise uninteresting subjects revealed in an entirely new way. Even better, it's pretty cheap to experiment with macro by picking up some inexpensive macro tubes that will work with your existing gear.

    DPR contributor Chris Foreman recently shared his experience getting back into macro photography, and detailed his unique approach using focus stacking and a motorized slider. Take a look at his images above for some inspiration and get all the details if you want to give it a try yourself.

  • Video: How Kodak became yet another victim of the juggernaut of changing technology

    The Eastman Kodak Company, better known as Kodak, was co-founded by George Eastman and Henry A. Strong way back in 1888. In the more than 130 years since, the company has been through a lot. It once was a dominant force in the photography industry, at times being almost as synonymous with photography as the Apple iPod once was with respect to music. For all but the youngest in society, we have all heard the phrase, 'Kodak moment.'

    However, with the advent of digital photography, Kodak began to lose its place at the forefront of photography. Interestingly, Kodak itself patented the first handheld digital camera in 1978, eventually shelving the product in order to better preserve its film business. With the benefit of hindsight, we can see that Kodak created its own demise a few decades in advance.

    Believing in the long-term viability of a predominantly analog photography industry, Kodak missed critical opportunities to diversify, a mistake Fujifilm didn't make, as is evidenced by the Japanese company's impressive lineup of digital photography cameras and lenses.

    Kodak's meteoric rise to near-total market dominance and the precipitous fall from grace is a fascinating story. The Wall Street Journal recently produced a short documentary covering the story of Kodak, which can be viewed below. The video begins with the origins of the company before ramping up to Kodak's prime and its eventual decline.

    In Kodak's heyday, the company focused on creating simple and easy-to-use still and video cameras for everyday people. This goal, one the company had long met successfully, resulted in Kodak employing more than 145,000 people in 1988. It was Kodak's best-ever year. At the time, Fujifilm was the biggest threat to Kodak's business. In 1989, Fujifilm managed to capture an 11% share of the U.S. film market. The next few years saw Kodak's position of power further eroded, and the company brought in a new CEO from outside in an attempt to right the slowly sinking ship.

    Kodak's digital camera shipments in the United States from 2008 through 2011. Image credit: The Wall Street Journal. Data credit: IDC.

    Losses continued to pile up and jobs were cut as Kodak moved to a new CEO at the turn of the 21st century. Kodak continued to produce new digital photography product and jumped into new product sectors, such as printers, but Kodak was fighting a losing battle. The advent of smartphone photography added further difficulties for Kodak, as no longer were consumers as inclined to buy a dedicated camera to capture life's most important moments. No longer was there a desire to create 'Kodak moments.'

    Kodak's stock prices have trended very downward since 2014. Image credit: Wall Street Journal. Data credit: Factset.

    Eventually, the financial challenges became too great and Kodak shuttered its camera business and filed for bankruptcy in 2012. However, the story of Kodak didn't end there. Desperate times call for desperate measures. Kodak began licensing its very identity, resulting in the Kodak name being plastered on many different products. Kodak even made a cryptocurrency in 2018. However, its stock price kept falling.

    With the ongoing coronavirus pandemic, Kodak found a new opportunity to pivot. On July 28, President Trump announced Kodak would receive a $765M government loan under the Defense Production Act. Kodak, a company long known for producing products using its chemical expertise, is poised to move into pharmaceuticals. Trump's announcement sent Kodak stock skyrocketing. However, the deal is currently being held up as the US Securities and Exchange Commission (SEC) investigates the deal and Kodak itself for insider trading.

    Kodak is far from the only company that has faltered in the digital age of photography. However, Kodak may be one of the most famous examples of how changing photographic technology can sink a company.