def init_weightsself superResNet selfinit_weights if isinstanceselfinit_cfg dict 报错RuntimeError Sizes of tensors must match except in dimension 2 Got 8 and 7 The offending ind

This is a code snippet for a ResNet model with a modified forward function. The forward function extracts patches from the input image, applies a Local Binary Pattern (LBP) layer to each patch, and then concatenates the LBP features with the output of each ResNet block. The resulting feature maps are returned as a tuple.

The init_weights function initializes the weights of the model. If the initialization configuration is set to use a pretrained model, the function returns without initializing the weights. Otherwise, it initializes the weights of each bottleneck block using Kaiming initialization and sets the output of the second batch normalization layer in each basic block to zero.

Note: window_size and stride are assumed to be global variables defined elsewhere in the code.