Tag: multi-scale feature fusion

Researchers report an improved YOLO model that detects small PCB surface defects with 90.63% sensitivity while reducing model size and ...

A lightweight YOLOv8-based detection model with three-branch attention and enhanced multi-scale fusion achieved 99.2 percent average precision on insulator defects ...

A CNN that fuses shallow and deep lesion features classified rice blast severity with about 93% accuracy and estimated the ...

AMA-YOLOv8s reached 93.5% mAP@0.5 in a dataset-based evaluation of photovoltaic defect detection.

A new probabilistic pose estimation framework combining multi-scale Vision Transformer feature fusion with adaptive uncertainty modeling improves accuracy on occluded ...

Chinese researchers have developed CDCNet, a deep learning framework that recognizes fine-grained barefoot footprint regions with 83.2 percent mean Average ...

Researchers in China have developed HA-DETR, a transformer-based detector that adaptively enhances features at different semantic levels to overcome blur, ...

Researchers in China have developed UG-HRNet, a lightweight high-resolution network that pairs a unified guidance deformable convolution with an alternating ...

Researchers in China have developed MAF-Net, a multi-scale attention fusion network that achieves a record Dice score of 85.47 percent ...

Researchers in China have developed HSF-Net, a hybrid-scale fusion deep learning network that achieved an 83.4 percent Dice score in ...