The following problems occurred when loading the weight file:
cudnn status error in: File: D: \Darknet master\SRC\revolutionary_layer.c : cudnn_convolutional_setup() : line: 253 : build time: Dec 31 2021 – 14:37:05

cuDNN Error: CUDNN_STATUS_BAD_PARAM

The online search said that the data type was wrong and there was not enough video memory, but they didn’t solve my problem

After tossing for a day, it was found that the parameters read when loading the cfg file did not match those in the file, as shown in the following figure:

By comparing the yolowrapper in Darknet with that in my program, it turns out that the two versions are different. There is one more initializeyolo parameter in the initialization function Use the yolowrapper in the program to solve the Darknet problem.