U-Net
PulseAugur coverage of U-Net — every cluster mentioning U-Net across labs, papers, and developer communities, ranked by signal.
- used by ResNet-34 90%
- used by GeoAI 90%
- used by Sentinel-2 90%
- instance of Brain Tumor Segmentation Using Convolutional Neural Networks in MRI Images 90%
- used by ResNet34 90%
- used by AlphaEarth 90%
- instance of Medical Image Segmentation 90%
- used by Sentinel-1 80%
- used by alphaXiv 70%
- used by Gotit.pub 70%
- used by ScienceCast 70%
- used by CatalyzeX 70%
14 day(s) with sentiment data
-
CMDiff: New Diffusion Model for Left Atrial Scar Segmentation
Researchers have developed CMDiff, a novel conditional mask diffusion model designed for segmenting left atrial scar tissue from 3D LGE MRI scans. This technique aims to improve the assessment of scar tissue, which is c…
-
New ALD-Font model enhances few-shot Chinese font generation
Researchers have developed a new model called ALD-Font for generating Chinese fonts with limited examples. This architecture-level disentangled diffusion model effectively separates content and style, improving both acc…
-
New Task-Performance Routing method enhances CT medical image segmentation
Researchers have developed a novel method called Task-Performance Routing (TPR) to improve CT medical image segmentation. This technique dynamically adjusts the weights of pre-trained models during training based on reg…
-
New differentiable watershed layer enhances AI image segmentation
Researchers have developed a novel differentiable segmentation layer, termed the "WS layer," based on the Power Watershed (PW) transform. This layer integrates classical morphological segmentation techniques with deep l…
-
New dataset enables ECG reconstruction from vibrational signals
Researchers have developed Vib2ECG, a new dataset and benchmark for reconstructing electrocardiography (ECG) signals from vibrational data. This approach aims to enable more accessible, mobile-device-friendly ECG monito…
-
New AI method improves spine landmark detection using contrastive learning
Researchers have developed a novel approach to spine landmark detection by reframing it as a pixel-level contrastive representation learning problem. This method uses dense pixel embeddings matched with landmark prototy…
-
U-VLM model advances automated radiology report generation
Researchers have developed U-VLM, a novel hierarchical vision-language model designed to improve automated radiology report generation from 3D medical imaging. Unlike previous models, U-VLM incorporates segmentation-pre…
-
AI models for quantum-dot recognition show poor transfer from simulation to experiment
A new research paper benchmarks five different neural network architectures for quantum-dot state recognition. The study found that models performing well on simulated data did not necessarily perform as well on experim…
-
Satellite bathymetry models show variable accuracy in Gulf waters
A new benchmark study evaluated satellite-derived bathymetry (SDB) models in the shallow, turbid waters of the Arabian Gulf. The research found that model performance varied significantly based on validation methods, wi…
-
New VarNet-Seg network improves prostate lesion MRI segmentation
Researchers have developed a new end-to-end network called VarNet-Seg for segmenting and detecting prostate lesions from undersampled diffusion-weighted MRI data. This network utilizes complex multicoil DWI data and coi…
-
AI model trained to detect illegal ivory carvings from images
Researchers have developed a machine learning model to automatically identify ivory carvings from images, aiming to combat the illegal wildlife trade. The proof-of-concept system utilizes a U-Net architecture with a Res…
-
SWIFT-Cell model offers faster, accurate cell segmentation
Researchers have developed SWIFT-Cell, a new segmentation model designed for faster cell identification in biological imaging. SWIFT-Cell is a distilled version of the Cellpose-SAM model, significantly reducing processi…
-
New Directional Attention Guidance speeds up diffusion models
Researchers have introduced Directional Attention Guidance (DAG), a novel training-free method to enhance the quality and speed of diffusion models. DAG addresses limitations of Classifier-Free Guidance (CFG) by analyzi…
-
New lightweight model offers energy-efficient lung tumor segmentation
Researchers have developed a new, energy-efficient model for segmenting lung tumors, crucial for cancer detection and treatment planning. This model, based on the ENet architecture and enhanced with U-Net style skip con…
-
New frameworks tackle LLM reliability and pathology segmentation · 2 sources tracked
Researchers have developed a new framework called Scoped Semantic Invariance to address inconsistencies in Large Language Model (LLM) responses. This approach formalizes consistency by evaluating model behavior within b…
-
Compact AI models show efficiency in flood mapping, competitive accuracy
Researchers have evaluated the effectiveness of compact foundation models, specifically Prithvi-EO-2.0-Tiny, for mapping surface water during floods. The study compared Prithvi-Tiny against a U-Net model and spectral-in…
-
New loss function enhances medical image segmentation accuracy
Researchers have developed a new loss function called Self-Tuned Boundary-Aware t-vMF Dice Loss (STB-tvMF Dice) to improve medical image segmentation. This method addresses challenges like class imbalance and boundary e…
-
New denoiser networks offer provable stability for image restoration
Researchers have developed new denoiser networks that offer provable stability guarantees, ensuring that small input perturbations result in only minimal output changes. These networks, designed by composing proximal la…
-
Efficient U-Net model developed for rapid spinal CT segmentation
Researchers have developed SpineContextResUNet, a computationally efficient 3D Residual U-Net designed for rapid spinal CT segmentation. This model addresses the high resource demands of current state-of-the-art methods…
-
ROCOCO model advances occluded object completion using flow matching
Researchers have developed ROCOCO, a novel generative model for completing occluded curvilinear objects, utilizing a flow matching-based approach. This method can inpaint occluded regions without prior knowledge of the …