Computer Vision

Computer vision is one of our deepest areas, covering how machines learn to see, segment, and reason about images and video. Articles here range from convolutional and transformer based architectures to dense prediction tasks like detection and segmentation, with regular coverage of medical imaging where reliable vision models carry real clinical weight. The emphasis stays on what makes a method work and where it breaks, backed by the original research.

PEGN: How Persistent Homology Breaks the WL Barrier in Graph Neural Networks.

PEGN: How Persistent Homology Breaks the WL Barrier in Graph Neural Networks

Leave a Comment / Computer Vision, Machine Learning /

PEGN: How Persistent Homology Breaks the WL Barrier in Graph Neural Networks | AI Trend Blend Graph Learning · Journal of Machine Learning Research 26 (2025) 1–36 · 20 min read Loops, Cycles, and the Topology GNNs Cannot See: How PEGN Breaks the Weisfeiler-Lehman Ceiling A multi-institution team spanning Peking University, UC San Diego, Stony […]

PEGN: How Persistent Homology Breaks the WL Barrier in Graph Neural Networks Read More »

Railway Sinkhole Detection with Physics-Informed Synthetic Data and SuperPoint Transformer.

Railway Sinkhole Detection with Physics-Informed Synthetic Data and SuperPoint Transformer

Leave a Comment / Machine Learning, Computer Vision, Remote Sensing AI, Vision Transformers & Attention /

Railway Sinkhole Detection with Physics-Informed Synthetic Data and SuperPoint Transformer | AI Trend Blend Infrastructure AI · ISPRS Journal of Photogrammetry and Remote Sensing 236 (2026) 487–499 · 21 min read How French Railway Engineers Taught an AI to Find Sinkholes It Had Almost Never Seen Before When you only have a handful of real

Railway Sinkhole Detection with Physics-Informed Synthetic Data and SuperPoint Transformer Read More »

RepVIS-GAN: Nighttime Satellite Visible Image Retrieval from Infrared Data.

RepVIS-GAN: Nighttime Satellite Visible Image Retrieval from Infrared Data

Leave a Comment / Machine Learning, Computer Vision, Generative & Diffusion Models, Remote Sensing AI /

RepVIS-GAN: Nighttime Satellite Visible Image Retrieval from Infrared Data | AI Trend Blend AITrendBlend Machine Learning Computer Vision About Satellite AI · ISPRS Journal of Photogrammetry and Remote Sensing 236 (2026) 162–174 · 20 min read RepVIS-GAN: Teaching a Satellite to See in the Dark by Reading the Heat It Can Already Feel Every night,

RepVIS-GAN: Nighttime Satellite Visible Image Retrieval from Infrared Data Read More »

SUP-Net: Deep Learning Fixes Doppler Ultrasound Aliasing by Upsampling the Raw Signal

SUP-Net: Deep Learning Fixes Doppler Ultrasound Aliasing by Upsampling the Raw Signal

Leave a Comment / Machine Learning, Computer Vision, Cybersecurity /

SUP-Net: Deep Learning Fixes Doppler Ultrasound Aliasing by Upsampling the Raw Signal | AI Trend Blend AITrendBlend Machine Learning Computer Vision About Medical AI · IEEE Transactions on Medical Imaging, Vol. 45, No. 1 (Jan 2026) · 18 min read The Aliasing Problem That Breaks Blood Flow Ultrasound — and How SUP-Net Solves It Without

SUP-Net: Deep Learning Fixes Doppler Ultrasound Aliasing by Upsampling the Raw Signal Read More »

H2CL: Dual-Geometry Hyperbolic-Euclidean Image-Text Learning for Medical Hierarchical Classification.

H2CL: Dual-Geometry Hyperbolic-Euclidean Image-Text Learning for Medical Hierarchical Classification

Leave a Comment / Machine Learning, Computer Vision, Medical AI, Multimodal AI /

H2CL: Dual-Geometry Hyperbolic-Euclidean Image-Text Learning for Medical Hierarchical Classification | AI Trend Blend AITrendBlend Machine Learning Computer Vision Medical AI About Medical AI · Medical Image Analysis, Vol. 112 (2026) · 20 min read Why Flat Classifiers Fail Doctors: H²CL Uses Hyperbolic Geometry to Teach AI the Clinical Hierarchy of Disease A UNSW Sydney team

H2CL: Dual-Geometry Hyperbolic-Euclidean Image-Text Learning for Medical Hierarchical Classification Read More »

GTP: The Graph-Transformer That Reads Whole Slide Pathology Images Like a Pathologist

GTP: The Graph-Transformer That Reads Whole Slide Pathology Images Like a Pathologist

Leave a Comment / Machine Learning, Computer Vision, Medical AI, Vision Transformers & Attention /

GTP: The Graph-Transformer That Reads Whole Slide Pathology Images Like a Pathologist | AI Trend Blend AITrendBlend Machine Learning Computer Vision Medical AI About Medical AI · IEEE Transactions on Medical Imaging, Vol. 41, Nov. 2022 · 22 min read GTP: The Model That Learned to Read Cancer Slides the Way a Pathologist Actually Does

GTP: The Graph-Transformer That Reads Whole Slide Pathology Images Like a Pathologist Read More »

Agentic AI for Personalized Knee Braces: How sEMG, Facial Expressions, and LLMs Combine to Configure Rehab Devices

Leave a Comment / Machine Learning, Agent Systems, Computer Vision, Medical AI /

Agentic AI for Personalized Knee Braces: How sEMG, Facial Expressions, and LLMs Combine to Configure Rehab Devices | AI Trend Blend AITrendBlend Machine Learning Computer Vision About Healthcare AI · Advanced Engineering Informatics 74 (2026) 104695 · 22 min read Your Knee Brace Said It Hurts. The AI Didn’t Believe It — Until the Muscle

Agentic AI for Personalized Knee Braces: How sEMG, Facial Expressions, and LLMs Combine to Configure Rehab Devices Read More »

PARNet: Dual-Encoder Crack Detection with Dynamic Alignment and Residual Fusion.

PARNet: Dual-Encoder Crack Detection with Dynamic Alignment and Residual Fusion

Leave a Comment / Machine Learning, Computer Vision, Engineering AI /

PARNet: Dual-Encoder Crack Detection with Dynamic Alignment and Residual Fusion | AI Trend Blend AITrendBlend Machine Learning Computer Vision About Computer Vision · Advanced Engineering Informatics 74 (2026) · Shandong University · 20 min read PARNet: The Crack Detection Network That Learned to See Like a Human Inspector — and Then Outperformed Eight of Them

PARNet: Dual-Encoder Crack Detection with Dynamic Alignment and Residual Fusion Read More »

MEWS: Semantic Segmentation With Almost No Labels — A Few Pixels Per Class Is All You Need.

MEWS: Semantic Segmentation With Almost No Labels — A Few Pixels Per Class Is All You Need

Leave a Comment / Machine Learning, Computer Vision, Image Segmentation /

MEWS: Semantic Segmentation With Almost No Labels — A Few Pixels Per Class Is All You Need | AI Trend Blend AITrendBlend Machine Learning Computer Vision Image Segmentation About Computer Vision · Neurocomputing 680 (2026) 133290 · 18 min read MEWS: The Segmentation Framework That Beats CLIP With Just a Few Pixel Clicks Per Class

MEWS: Semantic Segmentation With Almost No Labels — A Few Pixels Per Class Is All You Need Read More »

Mask-CDKD: Source-Free Knowledge Distillation from SAM for Satellite Onboard Land Cover Mapping.

Mask-CDKD: Source-Free Knowledge Distillation from SAM for Satellite Onboard Land Cover Mapping

Leave a Comment / Machine Learning, Computer Vision, Remote Sensing AI /

Mask-CDKD: Source-Free Knowledge Distillation from SAM for Satellite Onboard Land Cover Mapping | AI Trend Blend Satellite AI & Remote Sensing · ISPRS J. Photogramm. Remote Sens. 236 (2026) 1–21 · Wuhan University / Emory · 28 min read Teaching a Satellite to See the World Without Labels: How Mask-CDKD Squeezes SAM Into a 30M-Parameter

Mask-CDKD: Source-Free Knowledge Distillation from SAM for Satellite Onboard Land Cover Mapping Read More »