This project mainly conducts the experiment of two-dimensional object detection as well as three-dimensional object detection. We use convolutional neutral network as method and makes an improvement from the current method in object detection. Considering Yolo v4 has an excellent performance on both accuracy and speed, we try to improve its performance by changing the structure of the network on both two-dimensional dataset and three-dimensional dataset. Our main code of Yolo v4 is copied from https://github.com/bubbliiiing/yolov4-pytorch, while Complex Yolo v4 is copied from https://github.com/maudzung/Complex-YOLOv4-Pytorch.