Several remote sensing methods are used to conduct research on the solar system and numerous planets. Crater analysis may provide vital information about planets, such as their relative age. A model for crater detection will benefit data administration, data processing, and scientific investigation. Additionally, it will assist in the development of enhanced landing systems. Even now, enhanced computer capacity enables us to do deep learning operations. Multiple computer vision challenges demonstrate the effectiveness of neural network-based designs. In crater detection, however, neural network-based designs are not widely used. We utilize grayscale satellite images of Mars to train YOLOv4, YOLOv4-tiny, and SSD MobileNetV2 FPNLite to locate craters. Despite being trained for fewer steps, the YOLOv4-tiny model achieved an assessment mAP (0.50) of 88.36%. In contrast to SSD MobileNetV2 FPNLite, which took 6000 training steps to get the 78.59% mAP (0.50) score, the YOLOv4 model only required 2900 training steps to achieve the same result.