Answers are grounded in this paper and cite their sources.
Traditional methods of lithological classification often rely on the expertise of appraisers and the use of sophisticated measuring instruments. These methods are susceptible to staff experience and are time-consuming. To overcome these limitations, researchers have explored the use of rock images and intelligent algorithms to automatically identify rocks. However, models developed for automatic rock properties identification often require high-performance equipment that cannot be readily deployed on lightweight edge devices. To address this problem, we significantly extend our previous research and propose a method for automatic rock properties identification called SBR-EfficientViT. The method is based on an efficient vision converter and builds on our previous training framework. We also developed a training and application flow framework for the method, which can run with memory requirements of less than 720 MB and graphics memory of 1.6 GB. Furthermore, the proposed SBR-EfficientViT-M1 method achieves an impressive accuracy of 94.75%. • A ViT model (SBR-EfficientViT) is proposed to recognise rock images. • The connection structure of the network is redesigned based on EfficientViT. • The rock categories can be identified by rock images in a fast batch. • SBR-EfficientViT reduces the application limitations of rock recognition research.
No takes yet. Share an insight, caveat, or question.
Guo et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: