Tag: computer vision

A deep learning framework trained on orthophotos from the 2011 Great East Japan Earthquake detected damaged roads and bridges with ...

A CNN running on a host PC with Arduino-based servo actuation predicted steering angles with a validation MAE of 0.093 ...

Researchers used diffusion models to generate weather- and lighting-altered training videos with intact annotations, improving cross-dataset foreground segmentation from 0.76 ...

A geological-feature enhanced Cascade Mask R-CNN suppresses spectral confusion and class imbalance to recognize landslides in high-resolution satellite imagery, cutting ...

A newly published study describes a unified YOLOv11–EfficientViT neural network that jointly detects traffic lights and signs and recognizes traffic ...

An enhanced RT-DETR-ResNet18 framework raises detection accuracy for camouflaged soybean pests, reaching 77.4% mAP@0.5 across 12 pest categories.

A dual-path attention fusion network combining Swin Transformer and frozen DINOv2 features reached 99.24% accuracy on a leakage-screened hold-out test ...

A new AI framework combining attention-enhanced YOLOv8 and Bald Eagle Optimization achieved 98.94% accuracy in detecting transmission and distribution faults ...

Researchers developed a lightweight SAR building segmentation framework that achieves high accuracy while maintaining computational efficiency for urban mapping.

A new ensemble deep learning framework called EnPreNiNet, paired with a night image enhancement technique, achieved a reported mean Average ...