Unet-master2

  CN对图像进行像素级的分类，从而解决了语义级别的图像分割（semantic segmentation）问题。与经典的CNN在卷积层之后使用全连接层得到固定长度的特征向量进行分类（全联接层＋softmax输出）不同，FCN可以接受任意尺寸的输入图像，采用反卷积层对最后一个卷积层的feature map进行上采样, 使它恢复到输入图像相同的尺寸，从而可以对每个像素都产生了一个预测, 同时保留了原始输入图像中的空间信息, 最后在上采样的特征图上进行逐像素分类。(CN classifies images at the pixel level, thus resolving the problem of semantic segmentation at the semantic level. Unlike classical CNN, which uses full-connection layer to get fixed-length feature vectors after convolution layer for classification (full-connection layer + soft Max output), FCN can accept any size of input image, and uses deconvolution layer to sample feature map of the last convolution layer to restore it to the same size of input image, so that each pixel can be generated. At the same time, the spatial information of the original input image is retained. Finally, the pixel-by-pixel classification is carried out on the feature map sampled above.)