An image recognition method, an apparatus, an electronic device, and a storage medium. The method may include extracting input matrices from a target image through input channels of a convolutional layer in a convolutional neural network model, obtaining first and second result matrices by traversing input matrices and performing input transformation and convolution kernel transformation, respectively, on the traversed input matrices through the row transformation reconstruction, performing matrix multiplication reconstruction on the first and second result matrices to obtain to-be-transformed matrices, performing output transformation on the to-be-transformed matrices through the row transformation reconstruction to obtain output matrices, splicing the output matrices and outputting the spliced output matrices through output channels of the convolutional layer to obtain a convolution result, and obtaining an image recognition result based on the convolution result into a pooling layer in the convolutional neural network model to perform an image recognition on the target image.