The accurate segmentation of brain tumors from medical images is critical for diagnosis and treatment planning. However, traditional methods struggle with complex tumor shapes inconsistent image quality which leads to suboptimal results. To address this challenge, we propose multiple tasking Wasserstein Generative Adversarial Network U-shape Network++ (MWG-UNet++) by integrating a U-Net architecture enhanced transformer layers combined Networks (WGAN) data augmentation. proposed model called...

This article explores methods to accelerate distributed computation, focusing on its application in machine learning. It discusses two primary concepts: coded multiplication and data shuffling, along with a non-linear core Random Access Memory (RAM) approach, presenting new avenues for future research. The challenges developments of computation learning are examined, general discussion applications across various scientific fields. need systems capable handling massive volumes has led the...

Impact craters are the most common geomorphic unit on lunar surface, and there a large number of impact different sizes morphologies surface. Lunar key basis for geological studies, their study allows exploration Moon other planets. Therefore, this paper builds convolutional neural network YOLO V7_CBAM based V7 attention mechanism identifying craters. Based full-moon CCD images DEM provided by NASA, comparison test was conducted precision both higher than that V7, from 72.31% to 74.65%...

In astronomy, physics, climate modeling, geoscience, planetary science, and many other disciplines, the mass of data often comes from spherical sampling. Therefore, establishing an efficient distortion-free representation is essential. This paper introduces a novel (global) coordinate system that free singularity. Contrary to classical coordinates, such as Cartesian or polar systems, proposed naturally defined on surface. The basic idea this originated planar barycentric coordinates describe...

This paper proofs that one-stage object detection neural network of YoloV5 also performs well on lunar crater detection. The experiment is based CCD data provided by LROC camera carried NASA's Lunar Reconnaissance Orbiter (LRO), and gets a result nearly 75% test set without any preprocessing operation the original image.

The Lunar impact crater is one of the most remarkable geographic features on surface Lunar. size and shape are important for future landings resource development surface. Hence, detecting craters an essential research orientation. In this paper, we introduced two novel convolutional neural networks that combine YOLO v7 deformable convolution Deformable Convolutional Networks, called YOLOv7_DCN YOLOv7_DCNv2. With global CCD data provided by NASA, experiment has been conducted under same...

Download This Paper Open PDF in Browser Add to My Library Share: Permalink Using these links will ensure access this page indefinitely Copy URL DOI

Lineament is a unique geological structure. The study of Lunar lineament structure has great significance on understanding its history and evolution surface. However, the existing geographic feature extraction methods are not suitable for In this paper, new method proposed based improved-UNet++ YOLOv5. Firstly, dataset created containing lineaments CCD data from LROC. At same time residual blocks replaced with VGG in down sample part UNet++ adding attention block between each layer....

Due to the influence of speckle in synthetic aperture radar (SAR) image, statistical dependencies among neighboring pixels should be considered SAR image segmentation. The spatially adaptive weighted possibilistic c-means (SAW-PCM) clustering algorithm is proposed which spatial information introduced into PCM approach directly adjust membership. relationship between described through Markov random fields (MRF). To preserve detail images, directional neighborhood system set established....

This article aims to detect Mars craters for the geological research and planetary research. The main method is use deep learning identify on using TIR images. YOLOv5 chosen as basic neural network in this method. To improve efficiency, author chose a system that can automatically generate training dataset YOLO, results shows it feasible craters.

A new edge tracking method for cornea in optical coherence tomography anterior chamber images has been proposed the paper. The approach detects of outside first. Then, it fixes detected edge, which follows future knowledge cornea, to make sure that only real could be left. Finally, fits fixed by fourth order least squares. advantage this is fit even abnormal corneas.

This paper presents an automatic classification algorithm of Mars surface lineament structure based on Resnet50 in DEM (digital elevation model) data. work aims to reduce the time spent by planetary researchers collection samples, so as concentrate scientific research method avoids problems that traditional DTA terrain analysis) technology can only be used locally and judgment threshold is difficult set due large differences around planet. The highest accuracy crater 98.15%, dorsum 100%,...

This paper proposes an algorithm(system) to automatically mark and create a data set of Mars craters for deep learning based on the existing coordinate information. The main purpose this automatic processing system is solve error problem low efficiency caused by manual marking. To make such system, we need know something about planetary science, as way get full view planet, differences between DEM (Digital Elevation Model) image TIR (Thermal Infrared image) maps are projected.In paper,...

Computer aided diagnosis (CAD) of human liver and its tumor can provide information for both patients doctors help them in disease diagnosis, treatment, tracking, etc. However, manually done a medical reporting will cost plenty time need expertise to finish the work reduce errors. Automatic segmentation efficient way solve problem, but there still exist difficulties due CT image shape, like is little difference between healthy part diseased part. Our research focuses on computer-only...

Whole heart segmentation is widely used in the clinical diagnosis and treatment of cardiovascular diseases. The deep learning model can effectively recognize diseased tissue by CT image. With development imaging technology, data contained cardiac image becomes complex. traditional methods have problems slow processing speed low accuracy. This study proposes a semantic network based on learning. method introduces an enhanced attention mechanism TransUNet, optimizes preprocessing strategy. Our...

With the development of remote sensing technology, using images to do change detection is gaining more and attention. In this paper, a novel approach for unsupervised proposed. Firstly, SWT used multi-scale decomposition; then use treelet fusion reconstructed difference at every level; finally, map obtained through level set segmentation. The effectiveness our method validated by experiments on five different data sets from resources. By comparing with other state-of-art technologies,...

Abstract Lunar domes have always been one of the important windows to understand lunar volcanic activities, but traditional geological dome identification methods are costly. This study attempts establish an automatic method for through FSOD (Few-Shot Object Detection). Since our previous research has trying automatically identify hills using recognition algorithms, team will try Few-Shot Detection this time. In study, researchers first obtained coordinates from list known and intercepted...

In this paper, a process proposal is designed to automatically produce datasets for the surface features of remote sensing data. The article takes Mars linear structure as an example dataset production, and experiments are conducted with U-Net recognition network. results show that solution can effectively help relevant has good scalability.

The geographic feature of terrestrial planets is a critical and important reference that could help researchers have further understanding planetary history its evo-lution. Traditionally, the detection specific landforms their parameter extraction basically relies on manual marking, these types approach may consume lot labor time costs. On other side, with development convolutional neural networks (CNNs),it able to handle more complicated tasks such as object semantic segmentation high...