The present study proposes a deep learning model, named DeepSleepNet, for automatic sleep stage scoring based on raw single-channel EEG. Most of the existing methods rely hand-engineered features which require prior knowledge analysis. Only few them encode temporal information such as transition rules, is important identifying next stages, into extracted features. In proposed we utilize Convolutional Neural Networks to extract time-invariant features, and bidirectional-Long Short-Term Memory...

Compressed sensing magnetic resonance imaging (CS-MRI) enables fast acquisition, which is highly desirable for numerous clinical applications. This can not only reduce the scanning cost and ease patient burden, but also potentially motion artefacts effect of contrast washout, thus yielding better image quality. Different from parallel imaging-based MRI, utilizes multiple coils to simultaneously receive MR signals, CS-MRI breaks Nyquist-Shannon sampling barrier reconstruct MRI images with...

This paper proposes a practical approach to addressing limitations posed by use of single active electrodes in applications for sleep stage classification. Electroencephalography (EEG)-based characterizations progression contribute the diagnosis and monitoring many pathologies sleep. Several prior reports have explored ways automating analysis EEG reducing complexity data needed reliable discrimination stages order make it possible perform studies at lower cost home (rather than only...

In this paper, we propose a way of synthesizing realistic images directly with natural language description, which has many useful applications, e.g. intelligent image manipulation. We attempt to accomplish such synthesis: given source and target text our model synthesizes meet two requirements: 1) being while matching the description; 2) maintaining other features that are irrelevant description. The should be able disentangle semantic information from modalities (image text), generate new...

It's useful to automatically transform an image from its original form some synthetic (style, partial contents, etc.), while keeping the structure or semantics. We define this requirement as "image-to-image translation" problem, and propose a general approach achieve it, based on deep convolutional conditional generative adversarial networks (GANs), which has gained phenomenal success learn mapping images noise input since 2014. In work, we develop two step (unsupervised) learning method...

Intelligent indoor robotics is expected to rapidly gain importance in crucial areas of our modern society such as at-home health care and factories. Yet, existing mobile robots are limited their ability perceive respond dynamically evolving complex environments because inherently sensing computing resources that are, moreover, traded off against cruise time payload. To address these formidable challenges, here we propose intelligent metasurface (I2MR), where all relegated a centralized...

Translating information between text and image is a fundamental problem in artificial intelligence that connects natural language processing computer vision. In the past few years, performance caption generation has seen significant improvement through adoption of recurrent neural networks (RNN). Meanwhile, text-to-image begun to generate plausible images using datasets specific categories like birds flowers. We've even from multi-category such as Microsoft Common Objects Context (MSCOCO)...

Deep learning has enabled major advances in the fields of computer vision, natural language processing, and multimedia among many others. Developing a deep system is arduous complex, as it involves constructing neural network architectures, managing training/trained models, tuning optimization process, preprocessing organizing data, etc. TensorLayer versatile Python library that aims at helping researchers engineers efficiently develop systems. It offers rich abstractions for networks, model...

Fast Magnetic Resonance Imaging (MRI) is highly in demand for many clinical applications order to reduce the scanning cost and improve patient experience. This can also potentially increase image quality by reducing motion artefacts contrast washout. However, once an field of view desired resolution are chosen, minimum time normally determined requirement acquiring sufficient raw data meet Nyquist-Shannon sampling criteria. Compressive Sensing (CS) theory has been perfectly matched MRI...

Arbitrary-oriented object detection (AOOD) from optical remote sensing imagery has to correctly generate delicate oriented boundary boxes (OBBs) and meanwhile identify their specific categories. However, how make detectors learn parameters of OBBs, especially for the crucial orientation information, category complex background becomes a challenge task. Therefore, in this article, exploring better way guide detector parametric information novel one-stage anchor-free called Posterior Instance...

The development of general robotic systems capable manipulating in unstructured environments is a significant challenge. While Vision-Language Models(VLM) excel high-level commonsense reasoning, they lack the fine-grained 3D spatial understanding required for precise manipulation tasks. Fine-tuning VLM on datasets to create Vision-Language-Action Models(VLA) potential solution, but it hindered by high data collection costs and generalization issues. To address these challenges, we propose...

Achieving human-level dexterity in robots is a key objective the field of robotic manipulation. Recent advancements 3D-based imitation learning have shown promising results, providing an effective pathway to achieve this goal. However, obtaining high-quality 3D representations presents two problems: (1) quality point clouds captured by single-view camera significantly affected factors such as resolution, positioning, and occlusions caused dexterous hand; (2) global lack crucial contact...

Robotic research encounters a significant hurdle when it comes to the intricate task of grasping objects that come in various shapes, materials, and textures. Unlike many prior investigations heavily leaned on specialized point-cloud cameras or abundant RGB visual data gather 3D insights for object-grasping missions, this paper introduces pioneering approach called RGBGrasp. This method depends limited set views perceive surroundings containing transparent specular achieve accurate grasping....

The primary challenge in the development of large-scale artificial intelligence (AI) systems lies achieving scalable decision-making—extending AI models while maintaining sufficient performance. Existing research indicates that distributed can improve scalability by decomposing complex tasks and distributing them across collaborative nodes. However, previous technologies suffered from compromised real-world applicability due to massive requirement communication sampled data. Here we develop...

Deep learning methods can play a crucial role in anomaly detection, prediction, and supporting decision making for applications like personal health-care, pervasive body sensing, so on. However, current architecture of deep networks suffers the privacy issue that users need to give out their data model (typically hosted server or cluster on Cloud) training prediction. This problem is getting more severe those sensitive health-care medical (e.g., fMRI sensors measures EEG signals). In...

Perceiving and manipulating 3D articulated objects (e.g., cabinets, doors) in human environments is an important yet challenging task for future home-assistant robots. The space of exceptionally rich their myriad semantic categories, diverse shape geometry, complicated part functionality. Previous works mostly abstract kinematic structure with estimated joint parameters poses as the visual representations objects. In this paper, we propose object-centric actionable priors a novel...