Impact craters are formed due to continuous impacts on the surface of planetary bodies. Most recent deep learning-based crater detection methods treat craters as circular shapes, and less attention is paid to extracting the exact shapes of craters. Extracting precise shapes of the craters can be helpful for many advanced analyses, such as crater formation. This paper proposes a combination of unsupervised non-deep learning and semi-supervised deep learning approach to accurately extract shapes of the craters and detect missing craters from the existing catalog. In unsupervised non-deep learning, we have proposed an adaptive rim extraction algorithm to extract craters' shapes. In this adaptive rim extraction algorithm, we utilized the elevation profiles of DEMs and applied morphological operation on DEM-derived slopes to extract craters' shapes. The extracted shapes of the craters are used in semi-supervised deep learning to get the locations, size, and refined shapes. Further, the extracted shapes of the craters are utilized to improve the estimate of the craters' diameter, depth, and other morphological factors. The craters' shape, estimated diameter, and depth with other morphological factors will be publicly available.