Lightweight Mars surface detection algorithm based on improved Yolov7

Mars surface detection is a challenging task that takes into account environmental factors such as terrain complexity, ground flatness, rock distribution, geological features, and storms. In this paper, we address the complexity of rock distribution on the surface of Mars and the resource-constrained difficulties of intelligences in performing the task, and propose the Yolov7- MFGC model on top of Yolov7, which solves the problems of existing rock detection models in terms of large computational load and number of parameters, as well as resource constraints in a specific environment. In this paper, we use a homemade Mars surface rock dataset for training and testing, and the test results show that Yolov7-MFGC improves the accuracy by 1.7%, reduces the amount of parameters by 49%, shrinks the model volume by 49.6%, and reduces the GFLOPS by 70% compared with Yolov7. Therefore, it can be concluded that Yolov7-MFGC is able to better perform the Mars surface rock detection task.