CAT-Net: Hybrid AI Watches Workers to Spot Unsafe Activity in Real Time
Researchers in India have developed CAT-Net, a hybrid deep learning model combining convolutional networks, coordinate attention, and Transformer encoders to ...
Researchers in India have developed CAT-Net, a hybrid deep learning model combining convolutional networks, coordinate attention, and Transformer encoders to ...
Researchers combined YOLOv8s and Pix2Pix with ant colony and particle swarm optimization to stabilize wrinkle detection and realistic cosmetic skin ...
Researchers at Institute of Science Tokyo combined epipolar geometry and appearance similarity to keep person identities consistent across multiple cameras, ...
Researchers in China have developed an unsupervised tensor decomposition algorithm that detects subtle, low-contrast fabric defects while keeping false alarms ...
Researchers have developed a multiobject tracking framework combining a nonlinear adaptive Kalman filter, multi-scale appearance enhancement, and trajectory stitching that ...
Researchers have introduced Light-wDEF, a resource-efficient and fully reproducible framework that enabled a fair comparison of ten visual deepfake detectors ...
Researchers have built a modular AI pipeline that indexes hours of video and retrieves specific people from text, image, or ...
Researchers in China have developed HA-DETR, a transformer-based detector that adaptively enhances features at different semantic levels to overcome blur, ...
Researchers have developed DPAS-Net, a deep learning framework that maintains remarkably accurate multi-view clustering even when up to 90 percent ...
Researchers have developed a memory-augmented CNN-vision transformer autoencoder that uses Farneback optical flow to detect anomalies in surveillance video with ...
© 2026 Scienmag - Science Magazine
© 2026 Scienmag - Science Magazine