EfficientNet-B3
One of the EfficientNet series convolutional neural networks proposed by Google in 2019. It applies a compound scaling strategy to simultaneously expand network depth, width, and input resolution. EfficientNet-B3 uses a 300×300 input resolution, achieves approximately 81.6% ImageNet Top-1 accuracy, and maintains compact parameter count and computational cost.
Timeline (last 90 days)
The study found that EfficientNet-B3 can extract the most effective feature representations, generating embedding vectors of 1536 dimensions.
EfficientNet-B3 has an input resolution of 300×300 and achieves a Top-1 accuracy of approximately 81.6% on ImageNet.
The combination of EfficientNet-B3 and SVM achieved the strongest baseline performance, with an accuracy of 96.31% and a macro F1 score of 96.36%.
HyMLRaman introduces generative feature augmentation based on DDPM (Denoising Diffusion Probabilistic Model), generating synthetic features in the PCA-reduced EfficientNet-B3 latent space.
All Facts (4)
The study found that EfficientNet-B3 can extract the most effective feature representations, generating embedding vectors of 1536 dimensions.
50%UnverifiedEfficientNet-B3 has an input resolution of 300×300 and achieves a Top-1 accuracy of approximately 81.6% on ImageNet.
50%UnverifiedThe combination of EfficientNet-B3 and SVM achieved the strongest baseline performance, with an accuracy of 96.31% and a macro F1 score of 96.36%.
50%UnverifiedHyMLRaman introduces generative feature augmentation based on DDPM (Denoising Diffusion Probabilistic Model), generating synthetic features in the PCA-reduced EfficientNet-B3 latent space.
50%