源码三维LeNet-5网络的预训练神经网络工具箱模型

Posted 2021-04-03 MATLAB的科学与工程应用

我们实现的2D LeNet-5模型在其训练集上训练后，在灰度MNIST测试集上的准确率达到98.48%。为了将预先训练好的二维LeNet-5（MNIST）的可学习参数转换为三维LeNet-5，我们在三维空间中复制了二维滤波器（重复复制）。这是可能的，因为视频或3D图像可以被转换成图像切片序列。在训练过程中，我们期望3D LeNet-5在每一帧中学习模式。这个模型有大约26万个可学习的参数。

Our implementation of 2D LeNet-5 model achieved 98.48% accuracy on the grey-scale MNIST test set after training on its train set. To transfer the learnable parameters from pre-trained 2D LeNet-5 (MNIST) to 3D one, we duplicated 2D filters (copying them repeatedly) through the third dimension. This is possible since a video or a 3D image can be converted into a sequence of image slices. In the training process, we expect that the 3D LeNet-5 learns patterns in each frame. This model has about 260,000 learnable parameters.

simply, call "lenet5TL3Dfun()" function.

参考文献：

Ebrahimi, Amir, et al. “Introducing Transfer Learning to 3D ResNet-18 for Alzheimer’s Disease Detection on MRI Images.” 2020 35th International Conference on Image and Vision Computing New Zealand (IVCNZ), IEEE, 2020, doi:10.1109/ivcnz51579.2020.9290616.