This paper proposes two novel losses, Distance-IoU (DIoU) and Complete IoU (CIoU), for bounding box regression in object detection. DIoU loss is designed to minimize the normalized distance between the central points of predicted and target bounding boxes, leading to faster convergence compared to traditional IoU and GIoU losses. CIoU loss further incorporates three geometric factors: overlap area, central point distance, and aspect ratio, resulting in faster convergence and better performance. Both losses are integrated into state-of-the-art object detection algorithms such as YOLOv3, SSD, and Faster R-CNN, achieving significant improvements in detection accuracy. Additionally, DIoU is applied in non-maximum suppression (NMS) to enhance performance by considering both overlap area and central point distance. Experimental results on benchmark datasets like PASCAL VOC and MS COCO show that DIoU and CIoU losses outperform existing methods in terms of detection accuracy and convergence speed. The proposed losses are easy to integrate into existing detection frameworks and provide notable performance gains.This paper proposes two novel losses, Distance-IoU (DIoU) and Complete IoU (CIoU), for bounding box regression in object detection. DIoU loss is designed to minimize the normalized distance between the central points of predicted and target bounding boxes, leading to faster convergence compared to traditional IoU and GIoU losses. CIoU loss further incorporates three geometric factors: overlap area, central point distance, and aspect ratio, resulting in faster convergence and better performance. Both losses are integrated into state-of-the-art object detection algorithms such as YOLOv3, SSD, and Faster R-CNN, achieving significant improvements in detection accuracy. Additionally, DIoU is applied in non-maximum suppression (NMS) to enhance performance by considering both overlap area and central point distance. Experimental results on benchmark datasets like PASCAL VOC and MS COCO show that DIoU and CIoU losses outperform existing methods in terms of detection accuracy and convergence speed. The proposed losses are easy to integrate into existing detection frameworks and provide notable performance gains.