Swin Transformer
PulseAugur coverage of Swin Transformer — every cluster mentioning Swin Transformer across labs, papers, and developer communities, ranked by signal.
5 day(s) with sentiment data
-
CLFTv2: Efficient Camera-LiDAR Fusion for Autonomous Driving
Researchers have introduced CLFTv2, an advanced framework for semantic segmentation in autonomous driving that efficiently fuses camera and LiDAR data. This new model replaces global ViT attention with a Swin-based enco…
-
New framework learns object concepts from video motion without human labels
Researchers have developed a novel framework that learns object-centric visual representations from raw videos without human annotations or camera calibration. This approach leverages motion boundaries from optical flow…
-
New framework enhances image retrieval with combined manifold learning techniques
Researchers have developed a new framework for content-based image retrieval (CBIR) that combines projection-based and rank-based manifold learning strategies. This approach aggregates alternative low-dimensional featur…
-
Hybrid AI model achieves 99% accuracy in detecting GAN-generated faces
Researchers have developed a novel hybrid architecture that combines EfficientNet-B0's convolutional processing with a Swin Transformer backend for more efficient detection of GAN-generated synthetic faces. This new mod…
-
New SFG-SwinSR model enhances remote sensing image super-resolution
Researchers have developed a new model called SFG-SwinSR, which enhances remote sensing image super-resolution by incorporating a Spatial-Frequency Gated Feed-Forward Network. This novel approach distinguishes between l…
-
New framework adapts AI models for material recognition from sparse visual data
A new framework called Sparse Surface Understanding Framework (SSUF) has been developed to improve material recognition from incomplete visual data. SSUF adapts four pre-trained architectures—ConvAE, ViT, Swin Transform…
-
New AI framework enhances pediatric tumor diagnosis with multi-scale image analysis
Researchers have developed CoPath, a novel framework designed for accurate and lightweight diagnosis of peripheral neuroblastic tumors (pNTs) using pathological images. CoPath integrates CoHisNet, a multi-scale feature-…
-
New research explores advanced machine unlearning techniques for AI models · 4 sources tracked
Researchers are developing new methods for machine unlearning, the process of removing specific data or knowledge from AI models. One approach, Source-Free Class Relearning Audit (SFRA), focuses on recovering forgotten …
-
New methods enhance LoRA efficiency and stability for model adaptation
Researchers have developed two new methods to improve the efficiency and stability of Low-Rank Adaptation (LoRA) techniques used in parameter-efficient model adaptation. Normalized Low-Rank Adaptation (NoRA) normalizes …
-
Deep learning approach improves ovarian ultrasound classification accuracy
Researchers have developed a lesion-guided region-of-interest (ROI) deep learning approach for ovarian ultrasound classification, achieving high accuracy while reducing annotation effort. This method was evaluated on tw…
-
Iwin Transformer tackles ViT complexity with interleaved windows and convolution
Researchers have introduced the Iwin Transformer, a novel hierarchical vision transformer designed to overcome limitations in existing Vision Transformers (ViTs). This new architecture combines interleaved window attent…
-
New pruning method significantly reduces Vision Transformer computational costs
Researchers have developed a new method called Denoised Variance-Based Pruning with Optimal Brain Bias Compensation (DVBP + OB$^2$C) to reduce the computational overhead of Vision Transformers (ViTs). This technique use…
-
SCOUT framework enables direct semantic editing of face recognition templates
Researchers have introduced SCOUT, a novel framework designed to discover and directly manipulate semantic concepts within face recognition templates. This method utilizes mechanistic interpretability to learn sparse te…
-
Deep learning benchmark for hazelnut quality assessment using X-ray images
Researchers have developed a deep learning benchmark for classifying hazelnut quality using X-ray images. The study, which involved 799 segmented X-ray images, evaluated various single-model configurations and ensembles…
-
New mammography datasets and generalization framework improve AI accuracy
Researchers have introduced two new datasets, BreastMammo and DenseMammo, designed to improve AI model generalization in mammography across different clinical sites. They also proposed a domain generalization framework …
-
Brain MRI Foundation Models Primarily Encode Acquisition Site, Not Anatomy
Researchers have discovered that frozen foundation models, when used to represent brain MRI data, primarily encode the site where the MRI was acquired rather than anatomical or clinical information. This effect was obse…
-
Deep learning models for breast cancer detection benchmarked for performance and emissions
A new paper benchmarks seven deep learning models for breast cancer detection, evaluating their performance and environmental impact. The study found that while EfficientNet and ResNet offer strong accuracy, they also p…
-
New methods accelerate Vision Transformer adaptation for edge devices
Researchers have developed new methods for adapting Vision Transformers (ViTs) to specific tasks more efficiently. One approach uses genetic programming to evolve layer-specific scalar functions that approximate normali…
-
Swin Transformer and clinical data boost skin lesion classification accuracy
Researchers have developed a new multimodal framework for classifying skin lesions, aiming to improve early skin cancer diagnosis. This approach integrates visual features extracted by a Swin Transformer with structured…
-
CoInS-Net advances medical image analysis with joint interpolation and segmentation
Researchers have developed CoInS-Net, a novel network designed for the simultaneous interpolation and segmentation of medical images. This approach addresses limitations in existing methods that handle these tasks indep…