Deep learning

CRGenNet: Cloud-Free Optical Image Generation Using SAR and Contaminated Optical Data.

CRGenNet: Cloud-Free Optical Image Generation Using SAR and Contaminated Optical Data

Leave a Comment / Machine Learning, Computer Vision /

CRGenNet: Cloud-Free Optical Image Generation Using SAR and Contaminated Optical Data | AI Trend Blend AITrendBlend Machine Learning Computer Vision About Remote Sensing AI · ISPRS Journal of Photogrammetry and Remote Sensing 236 (2026) 255–272 · 22 min read CRGenNet: How Satellites Can See Through Clouds by Never Assuming the Sky Is Clear Researchers at […]

CRGenNet: Cloud-Free Optical Image Generation Using SAR and Contaminated Optical Data Read More »

HMHI-Net: Hierarchical Memory with Heterogeneous Interaction for Video Object Segmentation.

HMHI-Net: Hierarchical Memory with Heterogeneous Interaction for Video Object Segmentation

Leave a Comment / Machine Learning, Computer Vision /

HMHI-Net: Hierarchical Memory with Heterogeneous Interaction for Video Object Segmentation | AI Trend Blend AITrendBlend Machine Learning Computer Vision About Computer Vision · ACM Multimedia 2025 · arXiv:2507.22465 · 20 min read Shallow Features Matter: How HMHI-Net Fixes the Fundamental Flaw in Video Object Segmentation Memory Fudan University researchers discovered that every existing memory-based video

HMHI-Net: Hierarchical Memory with Heterogeneous Interaction for Video Object Segmentation Read More »

SAMM: SAM2 Fine-Tuned for Universal Material Micrograph Segmentation.

SAMM: SAM2 Fine-Tuned for Universal Material Micrograph Segmentation

Leave a Comment / Machine Learning, Image Segmentation /

SAMM: SAM2 Fine-Tuned for Universal Material Micrograph Segmentation | AI Trend Blend AITrendBlend Machine Learning Computer Vision About Materials Informatics · Advanced Powder Materials 5 (2026) 100404 · 20 min read SAMM: Teaching SAM2 to Read a Microstructure — and Generalise Across All of Materials Science Researchers at Central South University fine-tuned the Segment Anything

SAMM: SAM2 Fine-Tuned for Universal Material Micrograph Segmentation Read More »

PraNet-V2: Dual-Supervised Reverse Attention for Medical Image Segmentation.

PraNet-V2: Dual-Supervised Reverse Attention for Medical Image Segmentation

Leave a Comment / Machine Learning, Image Segmentation, Medical AI /

PraNet-V2: Dual-Supervised Reverse Attention for Medical Image Segmentation | AI Trend Blend AITrendBlend Machine Learning Computer Vision About Medical Computer Vision · Computational Visual Media (2026) · 18 min read PraNet-V2: How Dual-Supervised Reverse Attention Finally Fixes Background Blindness in Medical Segmentation Researchers at Nankai University tore apart the reverse attention mechanism they invented five

PraNet-V2: Dual-Supervised Reverse Attention for Medical Image Segmentation Read More »

BRAU-Net++: The Hybrid CNN-Transformer That Rethinks Sparse Attention for Medical Image Segmentation.

BRAU-Net++: U-Shaped Hybrid CNN-Transformer Network for Medical Image Segmentation

Leave a Comment / Machine Learning, Image Segmentation /

BRAU-Net++: U-Shaped Hybrid CNN-Transformer Network for Medical Image Segmentation | AI Trend Blend AITrendBlend Machine Learning Computer Vision About Medical Computer Vision · IEEE Transactions on Emerging Topics in Computational Intelligence (2024) · 22 min read BRAU-Net++: The Hybrid CNN-Transformer That Rethinks Sparse Attention for Medical Image Segmentation Researchers at Chongqing University of Technology built

BRAU-Net++: U-Shaped Hybrid CNN-Transformer Network for Medical Image Segmentation Read More »

MSBP-Net: The Lightweight Polyp Detector.

MSBP-Net: The Lightweight Polyp Detector That Learned to See Boundaries the Way Surgeons Do

Leave a Comment / Machine Learning, Image Segmentation /

MSBP-Net: The Lightweight Polyp Detector That Learned to See Boundaries the Way Surgeons Do AITrendBlend Machine Learning Medical AI About Medical Imaging · Pattern Recognition 170 (2026) 112101 · 20 min read The Polyp Segmenter That Sees What Colonoscopies Miss — and Does It in Real Time Researchers at Sichuan University of Science and Engineering

MSBP-Net: The Lightweight Polyp Detector That Learned to See Boundaries the Way Surgeons Do Read More »

Overview of the proposed FreDNet.

FreDNet: The Remote Sensing Segmenter That Learned to Hear the Image, Not Just See It

Leave a Comment / Machine Learning, Image Segmentation /

FreDNet: The Remote Sensing Segmenter That Learned to Hear the Image, Not Just See It AITrendBlend Computer Vision About Remote Sensing AI · IEEE Trans. Geoscience & Remote Sensing, Vol. 64, 2026 · 22 min read The Segmentation Model That Learned to Hear the Image, Not Just See It Researchers at Hohai University built a

FreDNet: The Remote Sensing Segmenter That Learned to Hear the Image, Not Just See It Read More »

A New Framework That Teaches Multi-Agent Systems to Spot Unreliable Sensors.

When Drones Learn to Distrust: A New Framework That Teaches Multi-Agent Systems to Spot Unreliable Sensors

Leave a Comment / Machine Learning, Agent Systems /

When Drones Learn to Distrust: A New Framework That Teaches Multi-Agent Systems to Spot Unreliable Sensors AITrendBlend Machine Learning About Multi-Agent Systems · Information Fusion 133 (2026) 104261 · 18 min read When Drones Learn to Distrust: The Sensor Fusion Framework That Teaches Multi-Agent Systems to Spot Bad Data in Real Time Researchers at the

When Drones Learn to Distrust: A New Framework That Teaches Multi-Agent Systems to Spot Unreliable Sensors Read More »

MaRI: The Structural Re-parameterization Breakthrough That Eliminated Redundant Computation in Kuaishou’s Ranking Models.

MaRI: How Kuaishou Solved the Hidden Redundancy Problem Plaguing Recommendation Models

Leave a Comment / Machine Learning, Recommender Systems /

MaRI: How Kuaishou Solved the Hidden Redundancy Problem Plaguing Recommendation Models | AI Systems Research AISecurity Research Machine Learning Cybersecurity About Recommendation Systems · arXiv:2602.23105v1 [cs.IR] · 14 min read MaRI: The Structural Re-parameterization Breakthrough That Eliminated Redundant Computation in Kuaishou’s Ranking Models How a team of researchers at Kuaishou discovered that the biggest bottleneck

MaRI: How Kuaishou Solved the Hidden Redundancy Problem Plaguing Recommendation Models Read More »

The WEMoE framework transforms critical MLP modules into dynamic mixture-of-experts structures while statically merging non-critical components. Input-dependent routing weights allow the model to adaptively blend task-specific knowledge, achieving superior multi-task performance over static merging methods.

WEMoE: How a Mixture-of-Experts Approach Is Solving the Multi-Task Model Merging Problem

Leave a Comment / Machine Learning /

WEMoE: How a Mixture-of-Experts Approach Is Solving the Multi-Task Model Merging Problem | MedAI Research MedAI Research Machine Learning About Deep Learning · TPAMI, 2026 · 18 min read The Static Model Merging Problem — and How WEMoE Learned to Adapt WEMoE introduces a dynamic mixture-of-experts approach to multi-task model merging, transforming how we combine

WEMoE: How a Mixture-of-Experts Approach Is Solving the Multi-Task Model Merging Problem Read More »

Follow by Email
Facebook
X (Twitter)
Tiktok