The growing use of deep learning in safety-critical applications, such as medical imaging, has raised concerns about limited labeled data, where this demand is amplified as model complexity increases, posing hurdles for domain experts to annotate data. In response to this, active learning (AL) is used to efficiently train models with limited annotation costs. In the context of deep neural networks (DNNs), AL often uses confidence or probability outputs as a score for selecting the most informative samples. However, modern DNNs exhibit unreliable confidence outputs, making calibration essential. We propose an AL framework that self-calibrates the confidence used for sample selection during the training process, referred to as Confident Active Learning with Integrated CalibratiOn (CALICO). CALICO incorporates the joint training of a classifier and an energy-based model, instead of the standard softmax-based classifier. This approach allows for simultaneous estimation of the input data distribution and the class probabilities during training, improving calibration without needing an additional labeled dataset. Experimental results showcase improved classification performance compared to a softmax-based classifier with fewer labeled samples. Furthermore, the calibration stability of the model is observed to depend on the prior class distribution of the data.

Perception is crucial in the realm of autonomous driving systems, where bird's eye view (BEV)-based architectures have recently reached state-of-the-art performance. The desirability of self-supervised representation learning stems from the expensive and laborious process of annotating 2D and 3D data. Although previous research has investigated pretraining methods for both LiDAR and camera-based 3D object detection, a unified pretraining framework for multimodal BEV perception is missing. In this study, we introduce CALICO, a novel framework that applies contrastive objectives to both LiDAR and camera backbones. Specifically, CALICO incorporates two stages: point-region contrast (PRC) and region-aware distillation (RAD). PRC better balances the region- and scene-level representation learning on the LiDAR modality and offers significant performance improvement compared to existing methods. RAD effectively achieves contrastive distillation on our self-trained teacher model. CALICO's efficacy is substantiated by extensive evaluations on 3D object detection and BEV map segmentation tasks, where it delivers significant performance improvements. Notably, CALICO outperforms the baseline method by 10.5% and 8.6% on NDS and mAP. Moreover, CALICO boosts the robustness of multimodal 3D object detection against adversarial attacks and corruption. Additionally, our framework can be tailored to different backbones and heads, positioning it as a promising approach for multimodal BEV perception.

Source: OpenAI's DALL·E 2 with prompt "a hyperrealistic picture of a robot reading the news on a laptop at a coffee shop" Welcome to the 2nd edition of Robo-Insight, a biweekly robotics news update! In this post, we are excited to share a range of remarkable advancements in the field, showcasing progress in hazard mapping, surface crawling, pump controls, adaptive gripping, surgery, health assistance, and mineral extraction. In the domain of hazard mapping, researchers have developed a collaborative scheme that utilizes both ground and aerial robots for hazard mapping of contaminated areas. The team improved the quality of density maps and lowered estimation errors by using a heterogeneous coverage control technique. In comparison to homogeneous alternatives, the strategy optimizes the deployment of robots based on each one's unique characteristics, producing better estimation values and shorter operation times.

We explore Calico, a miniature relocatable wearable system with fast and precise locomotion for on-body interaction, actuation and sensing. Calico consists of a two-wheel robot and an on-cloth track mechanism or "railway," on which the robot travels. The robot is self-contained, small in size, and has additional sensor expansion options. The track system allows the robot to move along the user's body and reach any predetermined location. It also includes rotational switches to enable complex routing options when diverging tracks are presented. We report the design and implementation of Calico with a series of technical evaluations for system performance. We then present a few application scenarios, and user studies to understand the potential of Calico as a dance trainer and also explore the qualitative perception of our scenarios to inform future research in this space.

Rich countries pour heart-stopping amounts of money into health care. Advanced economies typically spend about 10% of gdp on keeping their citizens in good nick, a share that is rising as populations age. The American system's heft and inertia, perpetuated by the drugmakers, pharmacies, insurers, hospitals and others that benefit from it, have long protected it from disruption. Its size and stodginess also explain why it is being covetously eyed by big tech. Few other industries offer a potential market large enough to move the needle for the trillion-dollar technology titans.

Google is betting that the future of healthcare is going to be structured data and AI. The company is applying AI to disease detection, new data infrastructure, and potentially insurance. In this report we explore Google's many healthcare initiatives and areas of potential future expansion. Google has always seen itself as more than a search and advertising company. Now it's turning its focus to healthcare, betting that its AI prowess can create a powerful new paradigm for the detection, diagnosis, and treatment of disease. "So tomorrow, if AI can shape healthcare, it has to work through the regulations of healthcare … In fact, I see that as one of the biggest areas is where the benefits will play out for the next 10 – 20 years." In short, Google seems to be going after the healthcare space from every possible angle. For example, did you know that Google has a project to release sterilized mosquitoes to control the spread of infectious disease? Or that the company has started a limited commercial rollout of its diabetes management program? Or that it appears to be exploring insurance? Note: For simplicity we use "Google" as shorthand for the larger Alphabet company, under which many of these healthcare initiatives fall. We explain the Alphabet structure below. As Google enters healthcare, it's leaning heavily on its expertise in AI. Health data is getting digitized and structured, from a new electronic record standard to imaging to DNA sequencing.

Google has made no secret of its overarching ambition to organise the world's information and make it accessible to anyone. And the healthcare industry has no shortage of such information, in any number of repositories and diverse formats, from MRI images to patient notes and data gathered from wearable devices. Google's DeepMind and the NHS: A glimpse of what AI means for the future of healthcare The Google subsidiary has struck a series of deals with organisations in the UK health service -- so what's really happening? Google has long sought to diversify its revenues streams away from search and advertising, the business it was founded on and which continues to make up the bulk of its revenue nearly 20 years later. So could health be the industry that helps the company to achieve that aim?

Tesla has solar projects on smaller islands, and Musk thinks they should be scalable to larger ones like Puerto Rico. Pixelbots swarming to show a rainbow dinosaur. All work and no play makes Jack a dull boy. The same goes for large businesses -- putting all of your company's effort into a single product can lead to stagnation and stale ideas. Exploring new ideas is one way to keep your company relevant for the long haul, even if the side business doesn't have much to do with your main operations. You're about to see some examples of this from large household names in the American business world.

America is not the great melting pot that poets like Ralph Waldo Emerson once extolled. At least, that's not the story that DNA tells, according to the genealogy company Ancestry. Using more than 770,000 spit samples taken from their customers over the last five years, its researchers mapped how people moved and married in post-colonial America. And their choices--especially the ones that kept communities apart--shaped today's modern genetic landscape. The study, published today in Nature Communications, combines a DNA database with family tree information collected over the company's 34-year history. "We're all living under the assumption that we are individual agents," says Catherine Ball, chief scientific officer at Ancestry and the leader of the study.

Calico, a company focused on aging research and therapeutics, today announced that Daphne Koller, Ph.D., is joining the company as Chief Computing Officer. In this newly created position, Dr. Koller will lead the company's computational biology efforts. She will build a team focused on developing powerful computational and machine learning tools for analyzing biological and medical data sets. She and her team will work closely with the biological scientists at Calico to design experiments and construct data sets that could provide a deeper understanding into the science of longevity and support the development of new interventions to extend healthy lifespan. Calico will try to use machine learning to understand the complex biological processes involved in aging.