Dual Branch Network Achieves Real Time Rail Sleeper Crack Detection
A lightweight dual-branch neural network combining convolutional and transformer architectures achieves real-time detection of rail sleeper cracks across challenging railway ...
A lightweight dual-branch neural network combining convolutional and transformer architectures achieves real-time detection of rail sleeper cracks across challenging railway ...
Researchers have developed a deep neural network architecture integrating four complementary attention modules to improve the detection of small unmanned ...
The EAM-DETR model utilizes enhanced feature extraction and multi-scale fusion to improve the accuracy and speed of automated surface defect ...
Researchers demonstrate that stochastic gradient descent reduces catastrophic forgetting while scene-conditioned tracking improves identity preservation in autonomous driving perception systems.
A systematic review of lane detection literature from 2017 to 2026 reveals that transformer and anchor-based models deliver state-of-the-art performance ...
The BCEA-YOLO framework improves underwater object detection by integrating multi-scale attention and adaptive feature fusion to effectively identify tiny, low-contrast ...
A lightweight detector named GFLNet improves mAP@0.5 over the baseline on the DIOR and UCAS-AOD datasets.
A new detection framework called TriFocus-YOLO improves fabric defect detection accuracy and maintains real-time speed by integrating edge enhancement and ...
A new lightweight RGB-IR fusion network enables real-time object detection on drones, achieving high accuracy in low-light conditions with significantly ...
Researchers report a lightweight detection model that improves small-drone recognition in complex backgrounds while sustaining real-time speeds on edge hardware.
© 2026 Scienmag - Science Magazine
© 2026 Scienmag - Science Magazine