Convolutional neural network (CNN) is a very effective method in object detection. YOLOv3-Tiny, which uses a CNN base, has excellent performance for object detection. The Winograd algorithm can reduce the use of large multiplication operations on the convolution layer. In this paper, we explore the reduction of hardware resources in the convolution layer of the YOLOv3-Tiny model using the Winograd algorithm of various sizes. The results of the calculation of this algorithm can reduce the use of multiplication operations 49.76% - 74.09%, which on the other hand, increases the number of addition operations,