Lifeboat Foundation

Early last year, our research team from the Visual Computing Group introduced Swin Transformer, a Transformer-based general-purpose computer vision architecture that for the first time beat convolutional neural networks on the important vision benchmark of COCO object detection and did so by a large margin. Convolutional neural networks (CNNs) have long been the architecture of choice for classifying images and detecting objects within them, among other key computer vision tasks. Swin Transformer offers an alternative. Leveraging the Transformer architecture’s adaptive computing capability, Swin can achieve higher accuracy. More importantly, Swin Transformer provides an opportunity to unify the architectures in computer vision and natural language processing (NLP), where the Transformer has been the dominant architecture for years and has benefited the field because of its ability to be scaled up.

So far, Swin Transformer has shown early signs of its potential as a strong backbone architecture for a variety of computer vision problems, powering the top entries of many important vision benchmarks such as COCO object detection, ADE20K semantic segmentation, and CelebA-HQ image generation. It has also been well-received by the computer vision research community, garnering the Marr Prize for best paper at the 2021 International Conference on Computer Vision (ICCV). Together with works such as CSWin, Focal Transformer, and CvT, also from teams within Microsoft, Swin is helping to demonstrate the Transformer architecture as a viable option for many vision challenges. However, we believe there’s much work ahead, and we’re on an adventurous journey to explore the full potential of Swin Transformer.

In the past few years, one of the most important discoveries in the field of NLP has been that scaling up model capacity can continually push the state of the art for various NLP tasks, and the larger the model, the better its ability to adapt to new tasks with very little or no training data. Can the same be achieved in computer vision, and if so, how?

With recent significant advances in brain implants, MailOnline talks to law professor Burkhard Schafer about how neurotechnologies could influence criminal trials in the future.

Astronomers build new telescopes and peer at the night sky to see what they might find. Janelia Group Leader Abraham Beyene takes a similar approach when looking at the cells that make up the human brain.

Beyene and his team design and synthesize new types of highly sensitive biosensors they use to peer at neurons to see what they can learn.

Continue reading “New nanosensor gives unprecedented look at dopamine release” »

Scientists have created “synthetic” mouse embryos from stem cells without a dad’s sperm or a mom’s egg or womb.

The lab-created embryos mirror a natural mouse embryo up to 8 ½ days after fertilization, containing the same structures, including one like a beating heart.

In the near term, researchers hope to use these so-called embryoids to better understand early stages of development and study mechanisms behind disease without the need for as many lab animals. The feat could also lay the foundation for creating synthetic human embryos for research in the future.

When the skull of Sahelanthropus tchadensis was discovered in Chad in 2001, it pushed back the age of the oldest known representative species of humanity by…

After a tantalizing year-and-a-half wait since the Mars Perseverance Rover touched down on our nearest planetary neighbor, new data is arriving—and bringing with it a few surprises.

The rover, which is about the size of car and carries seven scientific instruments, has been probing Mars’ 30-mile-wide Jezero crater, once the site of a lake and an ideal spot to search for evidence of ancient life and information about the planet’s geological and climatic past.

In a paper published today in the journal Science Advances, a research team led by UCLA and the University of Oslo reveals that rock layers beneath the crater’s floor, observed by the rover’s ground-penetrating radar instrument, are unexpectedly inclined. The slopes, thicknesses and shapes of the inclined sections suggest they were either formed by slowly cooling lava or deposited as sediments in the former lake.

Physicists at the Max Planck Institute of Quantum Optics have managed to entangle more than a dozen photons efficiently and in a defined way. They are thus creating a basis for a new type of quantum computer. Their study is published in Nature.

The phenomena of the quantum world, which often seem bizarre from the perspective of the common everyday world, have long since found their way into technology. For example, entanglement: a quantum-physical connection between particles that links them in a strange way over arbitrarily long distances. It can be used, for example, in a quantum computer—a computing machine that, unlike a conventional computer, can perform numerous mathematical operations simultaneously. However, in order to use a quantum computer profitably, a large number of entangled particles must work together. They are the basic elements for calculations, so-called qubits.

“Photons, the particles of light, are particularly well suited for this because they are robust by nature and easy to manipulate,” says Philip Thomas, a doctoral student at the Max Planck Institute of Quantum Optics (MPQ) in Garching near Munich. Together with colleagues from the Quantum Dynamics Division led by Prof. Gerhard Rempe, he has now succeeded in taking an important step towards making photons usable for technological applications such as quantum computing: For the first time, the team generated up to 14 entangled photons in a defined way and with high efficiency.

An algorithm developed by researchers from Helmholtz Munich, the Technical University of Munich (TUM) and its University Hospital rechts der Isar, the University Hospital Bonn (UKB) and the University of Bonn is able to learn independently across different medical institutions. The key feature is that it is self-learning, meaning it does not require extensive, time-consuming findings or markings by radiologists in the MRI images.

This federated algorithm was trained on more than 1,500 MRI scans of healthy study participants from four institutions while maintaining data privacy. The algorithm then was used to analyze more than 500 patient MRI scans to detect diseases such as multiple sclerosis, vascular disease, and various forms of brain tumors that the algorithm had never seen before. This opens up new possibilities for developing efficient AI-based federated algorithms that learn autonomously while protecting privacy. The study has now been published in the journal Nature Machine Intelligence.

Health care is currently being revolutionized by artificial intelligence. With precise AI solutions, doctors can be supported in diagnosis. However, such algorithms require a considerable amount of data and the associated radiological specialist findings for training. The creation of such a large, central database, however, places special demands on data protection. Additionally, the creation of the findings and annotations, for example the marking of tumors in an MRI image, is very time-consuming.

Clean water is essential for human survival. However, less than 3% of fresh water can be used as drinking water. According to a report published by the World Meteorological Organization, there is scarcity of drinking water for approximately 1 billion people worldwide, which is expected to rise to 1.4 billion by 2050.

Seawater desalination technology, which produces fresh water from seawater, could solve the problem of water scarcity. At the Korea Institute of Science and Technology (KIST), a research team led by Dr. Kyung Guen Song from the Center for Water Cycle Research, have developed a hybrid membrane distillation module that combines solar energy with hydrothermal heat pumps to reduce thermal energy consumption during the desalination process. Their results are published in Energy Conversion and Management.

Reverse osmosis and evaporation methods are relatively common seawater desalination processes; however, these methods can operate only at high pressures and temperatures. In comparison, the membrane distillation method produces fresh water by utilizing the vapor pressure generated by the temperature difference between the flowing raw water and treated water separated by a membrane. This approach has the advantage of low energy consumption, as fresh water can be generated at pressures of 0.2–0.8 bar, which is lower than atmospheric pressure, and temperatures of 50–60℃. However, large scale operation requires more thermal energy. Thus, research studies are required to reduce the use of thermal energy for commercial operation.