Head-to-head benchmark ranks EfficientNet-B0 top for spotting enamel caries in intraoral photos
A controlled benchmark of five CNN architectures on 2,000 intraoral photographs found EfficientNet-B0 led with 98.18 percent accuracy, though no ...
A controlled benchmark of five CNN architectures on 2,000 intraoral photographs found EfficientNet-B0 led with 98.18 percent accuracy, though no ...
A hybrid vision transformer framework called ViT-LGFNet combines convolutional local texture extraction with global self-attention, reaching 95.15% mean validation accuracy ...
A new lightweight three-branch neural network called FLIFuse classifies bacterial pneumonia, viral pneumonia, and normal chest X-rays with 91.4 percent ...
A controlled benchmark of 17 pretrained convolutional neural networks on 52,706 micro-CT rock slices found DenseNet architectures delivered the most ...
A controlled study of 1,196 coffee leaf images found that an explainable Vision Transformer classified nutrient deficiencies with up to ...
A stacking ensemble combining EfficientNet, ResNet and MobileNet classified healthy and diseased olive leaf images with 97.21% accuracy on a ...
A hybrid DeiT system reached 99.13% accuracy on a 38-class plant disease benchmark while combining local and global AI explanations ...
A lightweight Vision Transformer fine-tuned with AdamW and cosine scheduling classified Persian miniature paintings by school with 93.85% mean accuracy, ...
A 79,717-parameter attention-fused neural network called LeafXNet matches or beats far larger deep learning models at detecting watermelon, grape, and ...
Chinese researchers distilled three large neural networks into a lightweight model that identifies Ligustici Rhizoma et Radix decoction pieces and ...
© 2026 Scienmag - Science Magazine
© 2026 Scienmag - Science Magazine