在其卷积层使用堆叠的小卷积核代替大卷积核,不仅与原来具有相同的感受野,

Using stacked small convolution kernels instead of large convolution kernels in its convolution layer not only has the same receptive field as the original, but also reduces the parameters of the network, has a stronger learning ability for features, and better fits the features of the picture information. At the same time, in order to improve the convergence speed in the model training process, prevent over-fitting, improve the high-frequency detail information in the image, and protect the edge information of the image, etc., this paper improves the multi-loss fusion.

Using stacked small convolution kernels instead of large convolution kernels in its convolution layer can not only have the same receptive field as the original, but also reduce the parameters of the network, have stronger ability to learn features and better fit the feature information of the picture. At the same time, in order to improve the convergence speed in the process of model training, prevent over fitting, improve the high-frequency detail information in the image and protect the image edge information, this paper improves the multi loss fusion

Using stacked small convolution kernels instead of large convolution kernels in the convolution layer not only has the same receptive field as the original one, but also can reduce the parameters of the network, improve the learning ability of features and better fit the feature information of pictures. At the same time, in order to improve the convergence speed in the process of model training, prevent over-fitting, improve the high-frequency detail information in the image, protect the edge information of the image, etc., this paper improves the multi-loss fusion.