Tag: computer vision

A sine-guided Transformer–CNN network significantly outperformed 21 comparison methods on three benchmark datasets for salient-object detection in optical remote-sensing imagery.

A modified YOLOv8 detector improved conventional ship-detection accuracy in high-resolution SAR imagery, including on two external datasets, but reduced inference ...

A lightweight detection model achieved 90.6% mAP50 for potato leaf-disease localization while using fewer parameters and less computation than RT-DETR-R18.

RF-DETR Small detected trophallaxis states more accurately than YOLO26n in controlled Camponotus fellah footage, while the YOLO model processed frames ...

An OpenCV workflow measured Ti-6Al-4V fusion-zone grain sizes under four welding conditions, showing smaller grains with pulsing and ultrasonic vibration ...

GraphLane combined three-frame context, uncertainty-weighted graph reasoning and road-geometry constraints to improve lane-coordinate estimates in a CULane study, although its ...

An Android research prototype combines image processing, deep learning and internet-connected analysis to identify a person’s face type and display ...

The StreetTree preprint provides more than 12 million street-tree images from 133 countries and tests show that pretrained vision models ...

A geometry-aware mesh network achieved 80.46% accuracy across five dental occlusion classifications in a 200-scan proof-of-concept study using simulated quantum ...

A controlled indoor study found that a contour-guided NeRF and Mamba-based workflow can reconstruct and separate rice grains, support automated ...