TextStyleBrush is an AI research project that can copy the style of text in a photo using just a single word. It could one day unlock new potential for creative self-expression like personalized messaging and captions. The work will be submitted to a peer-reviewed journal.
Read more at ai.facebook.com
Facebook AI open-sourced the NetHack Learning Environment (NLE) last year. Winning a game of NetHack requires long term planning in an incredibly unforgiving environment. The competition runs from early June through October 15, and the winners will be announced in December.
Read more at ai.facebook.com
In robotics research, it is notoriously difficult to set up a perfectly reproducible experiment. Even in controlled robotic lab environments, real-world changes cause the baseline model to perform inconsistently. Google discusses how to address this fundamental challenge using random assignment and A/B testing.
Read more at ai.googleblog.com
Researchers have created a system to evaluate the balance and posture control of humanoid robots. The system can be used to compare the performance of robots with humans. The researchers used a humanoid robot called Lucy, which is 1.5m in height and weighs approximately 16.5 kgs.
Read more at techxplore.com
Researchers at UCLA develop a new holographic imaging technique. It can rapidly reconstruct microscopic images of samples with up to 50-fold acceleration. The technique is taking advantage of recurrent neural networks trained using deep learning. It incorporates spatial features from multiple holograms to digitally create holographic microscopy images.
Read more at techxplore.com
Research at Samsung appeared in the well-known journal Science Advances discussing a technology that bypasses the limitations of stretchable devices. A stretchable organic LED (OLED) display and a photoplethysmography (PPG) sensor can measure and display the user's heart rate in real time. This process allowed for an electronic stretchable device capable of assessing bodily metrics via the skin.
Read more at techxplore.com
Researchers have been analyzing cells using a special type of microscopy called Scanning Dielectric Force Volume Microscopy. They can create maps of an electrical physical property called the dielectric constant. Each of the biomolecules that make up cells has a different dielectic constant, so a map of this property is basically a. map of cell composition. The current gold standard optical method involves applying a fluorescent dye that can disrupt the cell being studied. Their approach doesn't require the addition of any potentially disruptive external agent.
Read more at techxplore.com
Facebook's AI models perform trillions of inference operations every day for billions of people. Adopting PyTorch as Facebook's default AI framework helps ensure that all the experiences will run optimally at Facebook scale. Over a year into the migration process, there are more than 1,700 Py Torch-based inference models in full production at Facebook. 93 percent of our new training models for identifying and analyzing content on Facebook are on PyTorch.
Read more at ai.facebook.com
Evaluating how well translation systems perform has been a major challenge for AI researchers. If researchers cannot measure or compare their results, they can't develop better translation systems. FLORES-101 is the missing piece, the tool that enables researchers to rapidly test and improve upon multilingual translation models like M2M-100. It will make it easy to translate from Bengali to Marathi as it is to translate English to Spanish.
Read more at ai.facebook.com
Layouts are commonly used to direct the order in which the reader parses a document to enable a better understanding. We use self-attention layers as building blocks of a Variational Transformer Network (VTN) model. The VTN model is able to extract meaningful relationships between the layout elements (paragraphs, tables, images, etc.), resulting in realistic synthetic documents. We show the effectiveness of this combination across different domains, such as scientific papers, UI layouts, and even furniture arrangements.
Read more at ai.googleblog.com
In recent years, self-supervised representation learning has significantly advanced due to the application of contrastive learning. However, little work has been done to successfully apply contrastive learn in the fully-super supervised domain. In Supervised Contrastive Learning, presented at NeurIPS 2020, we propose a novel loss function called SupCon. SupCon allows normalized embeddings from the same class to be pulled closer together, while embeddings from different classes are pushed apart.
Read more at ai.googleblog.com
Researchers have developed a framework that can create Android applications from text descriptions. Most existing models for creating software based on text descriptions are based on end-to-end neural machine translation (NMT) models. NMT models struggle to generate large programs with hundreds of lines of code. Researchers created a new formal language that captures the complexity of app source code.
Read more at in.news.yahoo.com
AMD and Samsung to develop RDNA 2 graphics technology for an Exynos mobile system-on-chip. The two companies aim to expedite graphics innovation in the mobile product market for all-around higher performance. Not many details have yet emerged as to the chip itself, nor the products in which it will be implemented. In the U.S., Samsung's flagship Galaxy phones come with Snapdragon SoCs from Qualcomm.
Read more at www.theverge.com
Researchers develop model that enables a drone to shoot a video based on a desired emotion or viewer reaction. Team from CMU, the University of Sao Paulo and Facebook AI Research tested their model by creating sample videos. The drone could also create videos that were exciting, calm, enjoyable or nerve-wracking, like a chase scene or someone dribbling a soccer ball.
Read more at techxplore.com
Scientists have fabricated a device that can mimic human brain cognitive actions. It is more efficient than conventional techniques in emulating artificial intelligence. Device mimics the human brain synapse to boost computational speed and power consumption efficiency. The human brain comprises nearly a hundred billion neurons consisting of axons and dendrites.
Read more at www.msn.com
Researchers from the University of Bristol have demonstrated how a new special type of camera can build a pictorial map of where it has been and use this map to know where it currently is. Knowing where you are on a map is one of the most useful pieces of information when navigating journeys. This is essential for smart devices from robot vacuum cleaners to delivery drones to wearable sensors.
Read more at techxplore.com
Researchers at Tokyo Institute of Technology and RIKEN, Japan, have performed a meta-synthesis to understand how we perceive and interact with the voice (and the body) of various machines. Their findings have generated insights into human preferences, and can be used by engineers and designers to develop future vocal technologies.
Read more at techxplore.com
Machine learning is the process by which computers adapt their responses without human intervention. A challenge posed by the rapid expansion of machine learning is high energy demand of complex computing processes. Researchers from The University of Tokyo have reported the first integration of a mobility-enhanced FET and a ferroelectric capacitor. Their findings were presented at the 2021 Symposium on VLSI Technology.
Read more at techxplore.com
Fuchsia, Google's from-scratch operating system, is now running on the first-generation Nest Hub. The update will not change any of the functionality of the Nest Hub's features. Under the hood, the smart display will be running Fuchsia OS instead of the Linux-based Cast OS. It's likely that Nest Hub owners won't even notice they've been switched over to Fuchsia OS, Google says.
Read more at 9to5google.com
Researchers studied 25 million self-referencing tweets from people in U.S. and Norway. They found that people's moods change based on how happy or unhappy they were during different parts of the day. The tweeters seemed to be less happy on Sunday afternoons and remained unhappy until Thursday. They also noticed that as a whole, the tweeters became less happy until around the following Thursday.
Read more at techxplore.com
