Accelerating Inference Of Cnns With In-Memory Computing