Hasil untuk "Maps"

Qibin Hou, Daquan Zhou, Jiashi Feng

Recent studies on mobile network design have demonstrated the remarkable effectiveness of channel attention (e.g., the Squeeze-and-Excitation attention) for lifting model performance, but they generally neglect the positional information, which is important for generating spatially selective attention maps. In this paper, we propose a novel attention mechanism for mobile networks by embedding positional information into channel attention, which we call "coordinate attention". Unlike channel attention that transforms a feature tensor to a single feature vector via 2D global pooling, the coordinate attention factorizes channel attention into two 1D feature encoding processes that aggregate features along the two spatial directions, respectively. In this way, long-range dependencies can be captured along one spatial direction and meanwhile precise positional information can be preserved along the other spatial direction. The resulting feature maps are then encoded separately into a pair of direction-aware and position-sensitive attention maps that can be complementarily applied to the input feature map to augment the representations of the objects of interest. Our coordinate attention is simple and can be flexibly plugged into classic mobile networks, such as MobileNetV2, MobileNeXt, and EfficientNet with nearly no computational overhead. Extensive experiments demonstrate that our coordinate attention is not only beneficial to ImageNet classification but more interestingly, behaves better in down-stream tasks, such as object detection and semantic segmentation. Code is available at https://github.com/Andrew-Qibin/CoordAttention.

Sanghyun Woo, Jongchan Park, Joon-Young Lee et al.

We propose Convolutional Block Attention Module (CBAM), a simple yet effective attention module for feed-forward convolutional neural networks. Given an intermediate feature map, our module sequentially infers attention maps along two separate dimensions, channel and spatial, then the attention maps are multiplied to the input feature map for adaptive feature refinement. Because CBAM is a lightweight and general module, it can be integrated into any CNN architectures seamlessly with negligible overheads and is end-to-end trainable along with base CNNs. We validate our CBAM through extensive experiments on ImageNet-1K, MS COCO detection, and VOC 2007 detection datasets. Our experiments show consistent improvements in classification and detection performances with various models, demonstrating the wide applicability of CBAM. The code and models will be publicly available.

Wenzhe Shi, Jose Caballero, Ferenc Huszár et al.

Recently, several models based on deep neural networks have achieved great success in terms of both reconstruction accuracy and computational performance for single image super-resolution. In these methods, the low resolution (LR) input image is upscaled to the high resolution (HR) space using a single filter, commonly bicubic interpolation, before reconstruction. This means that the super-resolution (SR) operation is performed in HR space. We demonstrate that this is sub-optimal and adds computational complexity. In this paper, we present the first convolutional neural network (CNN) capable of real-time SR of 1080p videos on a single K2 GPU. To achieve this, we propose a novel CNN architecture where the feature maps are extracted in the LR space. In addition, we introduce an efficient sub-pixel convolution layer which learns an array of upscaling filters to upscale the final LR feature maps into the HR output. By doing so, we effectively replace the handcrafted bicubic filter in the SR pipeline with more complex upscaling filters specifically trained for each feature map, whilst also reducing the computational complexity of the overall SR operation. We evaluate the proposed approach using images and videos from publicly available datasets and show that it performs significantly better (+0.15dB on Images and +0.39dB on Videos) and is an order of magnitude faster than previous CNN-based methods.

P. Ade, N. Aghanim, M. Arnaud et al.

This paper describes the mapmaking procedure applied to Planck Low Frequency Instrument (LFI) data. The mapmaking step takes as input the calibrated timelines and pointing information. The main products are sky maps of I , Q , and U Stokes components. For the first time, we present polarization maps at LFI frequencies. The mapmaking algorithm is based on a destriping technique, which is enhanced with a noise prior. The Galactic region is masked to reduce errors arising from bandpass mismatch and high signal gradients. We apply horn-uniform radiometer weights to reduce the e ff ects of beam-shape mismatch. The algorithm is the same as used for the 2013 release, apart from small changes in parameter settings. We validate the procedure through simulations. Special emphasis is put on the control of systematics, which is particularly important for accurate polarization analysis. We also produce low-resolution versions of the maps and corresponding noise covariance matrices. These serve as input in later analysis steps and parameter estimation. The noise covariance matrices are validated through noise Monte Carlo simulations. The residual noise in the map products is characterized through analysis of half-ring maps, noise covariance matrices, and simulations.

Max Jaderberg, K. Simonyan, Andrew Zisserman et al.

Convolutional Neural Networks define an exceptionally powerful class of models, but are still limited by the lack of ability to be spatially invariant to the input data in a computationally and parameter efficient manner. In this work we introduce a new learnable module, the Spatial Transformer, which explicitly allows the spatial manipulation of data within the network. This differentiable module can be inserted into existing convolutional architectures, giving neural networks the ability to actively spatially transform feature maps, conditional on the feature map itself, without any extra training supervision or modification to the optimisation process. We show that the use of spatial transformers results in models which learn invariance to translation, scale, rotation and more generic warping, resulting in state-of-the-art performance on several benchmarks, and for a number of classes of transformations.

Suhas S. P. Rao, M. Huntley, Neva C. Durand et al.

Jakob J. Engel, Thomas Schöps, D. Cremers

Min Lin, Qiang Chen, Shuicheng Yan

We propose a novel deep network structure called "Network In Network" (NIN) to enhance model discriminability for local patches within the receptive field. The conventional convolutional layer uses linear filters followed by a nonlinear activation function to scan the input. Instead, we build micro neural networks with more complex structures to abstract the data within the receptive field. We instantiate the micro neural network with a multilayer perceptron, which is a potent function approximator. The feature maps are obtained by sliding the micro networks over the input in a similar manner as CNN; they are then fed into the next layer. Deep NIN can be implemented by stacking mutiple of the above described structure. With enhanced local modeling via the micro network, we are able to utilize global average pooling over feature maps in the classification layer, which is easier to interpret and less prone to overfitting than traditional fully connected layers. We demonstrated the state-of-the-art classification performances with NIN on CIFAR-10 and CIFAR-100, and reasonable performances on SVHN and MNIST datasets.

C. Alvares, J. Stape, P. Sentelhas et al.

Koppen's climate classification remains the most widely used system by geographical and climatological societies across the world, with well recognized simple rules and climate symbol letters. In Brazil, climatology has been studied for more than 140 years, and among the many proposed methods Koppen 0 s system remains as the most utilized. Considering Koppen's climate classification importance for Brazil (geography, biology, ecology, meteorology, hydrology, agronomy, forestry and environmental sciences), we developed a geographical information system to identify Koppen's climate types based on monthly temperature and rainfall data from 2,950 weather stations. Temperature maps were spatially described using multivariate equations that took into account the geographical coordinates and altitude; and the map resolution (100 m) was similar to the digital elevation model derived from Shuttle Radar Topography Mission. Patterns of rainfall were interpolated using kriging, with the same resolution of temperature maps. The final climate map obtained for Brazil (851,487,700 ha) has a high spatial resolution (1 ha) which allows to observe the climatic variations at the landscape level. The results are presented as maps, graphs, diagrams and tables, allowing users to interpret the occurrence of climate types in Brazil. The zones and climate types are referenced to the most important mountains, plateaus and depressions, geographical landmarks, rivers and watersheds and major cities across the country making the information accessible to all levels of users. The climate map not only showed that the A, B and C zones represent approximately 81%, 5% and 14% of the country but also allowed the identification of Koppen's climates types never reported before in Brazil.

Peter Schad, S. Dondeyne

S. Hirtle, J. Jonides

Chao Lian, Chao Lian, Ling Liu et al.

Cancer immunotherapy has transformed oncology, yet its clinical efficacy is often limited by immune evasion within the tumor microenvironment (TME). Regulatory T cells (Tregs), a key immunosuppressive lineage, potently inhibit effector T-cell proliferation and activation, thereby dampening antitumor immune responses. Tregs are frequently enriched in diverse solid tumors, and their abundance correlates with poor prognosis, increased tumor invasiveness, and therapeutic resistance. A major mechanism driving this enrichment is the chemokine-chemokine receptor axis. Tumor cells, along with other stromal and immune cells in the TME, secrete chemokines including CCL22, CCL20, and CXCL12, which bind to CCR4, CCR6, and CXCR4 on Tregs and direct their recruitment and activation within the TME. This establishes an immunosuppressive niche that promotes tumor growth, facilitates metastasis, and reduces responsiveness to immunotherapy. This review consolidates eight experimentally validated chemokine-Treg axes from 2005 to 2025, with each study annotated by tumor type and represented by the highest observed level of evidence. A systematic representation illustrates how these axes mediate Treg-driven immunosuppression and maps their prevalence across cancers. Focusing on these axes provides mechanistic insights, highlights potential therapeutic targets, and identifies predictive biomarkers. Strategies targeting the chemokine-chemokine receptor axes, including selective receptor blockade, combination with immune checkpoint inhibitors, and omics-based approaches to resolve Treg heterogeneity, offer avenues to reprogram the immunosuppressive TME and enhance antitumor immunity.

Pavlos Kassotakis

We provide necessary and sufficient conditions for maps that satisfy associative-like conditions on families of n-ary magmas to be pentagon maps. We obtain parametric-pentagon maps and we propose a procedure that generates families of multicomponent pentagon and entwining pentagon maps from a given pentagon map.

Andrew B. Newman, Mahdi Qezlou, Gwen C. Rudie et al.

The Ly α Tomography IMACS Survey (LATIS) has produced large 3D maps of the intergalactic medium (IGM), providing a new window on the cosmic web at z ∼ 2.5. A key advantage of Ly α tomography is that it enables the discovery of overdense regions without the need to detect their galaxy members in spectroscopic surveys, circumventing possible selection biases. We use these maps to identify 37 IGM-selected overdensities as regions of strong and spatially coherent Ly α absorption. Simulations indicate that 85% of these are protoclusters, defined as the progenitors of z = 0 halos with mass M _desc > 10 ^14 M _⊙ , and that nearly all of the rest are protogroups (10 ^13.5 < M _desc / M _⊙ < 10 ^14 ). We estimate the masses and space densities of the IGM-selected overdensities and show they are in accordance with mock surveys. We investigate the LATIS counterparts of some previously reported protoclusters, including the proto-supercluster Hyperion. We identify a new component of Hyperion beyond its previously known extent. We show that the Ly α transmission of the galaxy density peaks within Hyperion is consistent with a simple physical model (the fluctuating Gunn–Peterson approximation), suggesting that active galactic nucleus feedback or other processes have not affected the large-scale gas ionization within this structure as a whole. The LATIS catalog represents an order-of-magnitude increase in the number of IGM-selected protogroups and protoclusters and will enable new investigations of the connections between galaxies and their large-scale environments at cosmic noon.

Muhammad Afaq Hussain, Zhanlong Chen, Biswajeet Pradhan et al.

Study region: The National Highways 85 and 50, key routes of the China–Pakistan Economic Corridor (CPEC) in Balochistan, Pakistan. Study focus: Flooding is a natural disaster that is becoming increasingly frequent and severe. The National Highways 85 and 50 are vulnerable, necessitating accurate flood susceptibility mapping (FSM). Current machine learning (ML) models for FSM often suffer from low efficiency and overfitting. This study introduces an innovative hybrid FSM approach using four heterogeneous ensemble learning (HEL) techniques combined with three ML models: Random Forest (RF), Support Vector Machine (SVM), and Light Gradient Boosting Machine (LGBM). The proposed method was tested using satellite data from Sentinel-1, Sentinel-2, and Landsat-8, analyzing 1371 flood locations and 12 contributing variables. RF, variable importance factors (VIF), and information gain ratio (IGR) were applied to assess multicollinearity. The dataset was split (70:30) for model training and testing, with HEL-based models achieving superior performance over single ML models. New hydrological insights for the region: The stacking model yielded the highest AUROC (0.98), Kappa (0.82), accuracy (0.927), precision (0.963), Matthew’s correlation coefficient (0.820), and F1-score (0.950). HEL-based models proved more stable and resistant to overfitting. IGR analysis identified slope and distance from streams as key factors in FSM. The resulting flood-prone maps provide insights for disaster management adaptation strategies, demonstrating the broader applicability of the developed approach to enhance FSM accuracy and reliability.

Rob Felix, Jaap Bouwman, Baudewijn Odé et al.

This paper presents all available information on the Orthoptera of the Socotra Archipelago, an area well-known for its endemic flora and fauna. General information is provided about the climate and geology of the Socotra Archipelago. The various habitats where grasshoppers have been found are described and illustrated, followed by a concise history of Orthoptera research on Socotra. Besides an identification key to the species, additional information about the material examined, taxonomy, diagnostic notes, distribution and occurrence, including maps, habitat, biology and bioacoustics, is provided for each species. In total, 65 Orthoptera species are reported here from Socotra, Abd el Kuri, Samha and Darsa, including Oecanthus castaneus Felix &amp;amp; Bouwman, sp. nov. and two unknown species assigned to Ectatoderus. Of these 65 species, 30 (46%) are endemic to the Socotra Archipelago. Re-descriptive notes on Acrotylus innotatus Uvarov, 1933 and Glomeremus capitatus Uvarov, 1957 are provided, including the description of the female of the latter species and the male of Oxytruxalis ensis (Burr, 1899). Acrotylus innotatus Uvarov, 1933, Dictyophorus griseus (Reiche &amp;amp; Fairmaire, 1850), Eumodicogryllus chivensis (Tarbinsky, 1930), Ochrilidia geniculata (Bolívar, 1913), Sphingonotus rubescens (Walker, 1870) and S. balteatus (Serville, 1838) are recorded for the first time from the Archipelago. Bioacoustics are presented for: Ochrilidia socotrae Massa, 2009, Stenohippus socotranus (Popov, 1957), Sphingonotus ganglbaueri Krauss, 1907, S. insularis (Popov, 1957), Acheta rufopictus Uvarov, 1957, Eumodicogryllus chivensis (Tarbinsky, 1930), Ectatoderus guichardi Gorochov, 1993 as well as two other species assigned to Ectatoderus, Oecanthus castaneus Felix &amp;amp; Bouwman, sp. nov., Ruspolia aff. R. basiguttata (Bolívar, 1906) and Pachysmopoda abbreviata (Taschenberg, 1883). Red List Assessments for 29 endemic species have been prepared including Oxytruxalis ensis (Burr, 1899) (Critically Endangered, CR), Cataloipus brunneri (Kirby, 1910) (Endangered, EN) and Glomeremus capitatus Uvarov, 1957, Phaneroptila insularis Uvarov, 1957, Phaulotypus granti Burr, 1899, Socotracris kleukersi Felix &amp;amp; Desutter-Grandcolas, 2012, Socotrella monstrosa Popov, 1957 and Xenephias socotranus Kevan, 1973 (all Vulnerable, VU).

András Csépai, András Szűcs

We introduce a new notion, called quasi-holomorphic maps. These are real smooth maps equipped with a structure that imitates the singularities and singularity stratifications of holomorphic maps on the source and target manifolds, although the manifolds themselves carry no global complex structures. Some important examples of quasi-holomorphic maps are branched coverings and links of finitely determined holomorphic map germs. We show a Pontryagin--Thom type construction for a ``universal'' quasi-holomorphic map with prescribed multisingularities, from which all such maps can be induced, and a similar result for maps with prescribed singularities. Applying this, we prove that the Thom polynomials of holomorphic singularities determine the cohomology classes represented by the singular loci of not only holomorphic but quasi-holomorphic maps as well. As another application we define the cobordism groups of quasi-holomorphic maps with restricted multisingularities, whose classifying space was given by the above construction. We completely compute the free parts of these cobordism groups and in some special cases also obtain results on their torsion parts.

D. Cox, Margit Burmeister, Elissa R Price et al.

J. Meiss

Maali Alabdulhafith, Abduljabbar S. Ba Mahel, Nagwan Abdel Samee et al.

Quality of life is greatly affected by chronic wounds. It requires more intensive care than acute wounds. Schedule follow-up appointments with their doctor to track healing. Good wound treatment promotes healing and fewer problems. Wound care requires precise and reliable wound measurement to optimize patient treatment and outcomes according to evidence-based best practices. Images are used to objectively assess wound state by quantifying key healing parameters. Nevertheless, the robust segmentation of wound images is complex because of the high diversity of wound types and imaging conditions. This study proposes and evaluates a novel hybrid model developed for wound segmentation in medical images. The model combines advanced deep learning techniques with traditional image processing methods to improve the accuracy and reliability of wound segmentation. The main objective is to overcome the limitations of existing segmentation methods (UNet) by leveraging the combined advantages of both paradigms. In our investigation, we introduced a hybrid model architecture, wherein a ResNet34 is utilized as the encoder, and a UNet is employed as the decoder. The combination of ResNet34’s deep representation learning and UNet’s efficient feature extraction yields notable benefits. The architectural design successfully integrated high-level and low-level features, enabling the generation of segmentation maps with high precision and accuracy. Following the implementation of our model to the actual data, we were able to determine the following values for the Intersection over Union (IOU), Dice score, and accuracy: 0.973, 0.986, and 0.9736, respectively. According to the achieved results, the proposed method is more precise and accurate than the current state-of-the-art.