Through a comprehensive evaluation of model complexity and number of parameters, it was determined that the overall performance of the proposed model is the best when eight group convolutions are used ...