arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2407.18100cs.CVphysics.geo-ph
收藏

DINOv2 助力地质图像分析:分类、分割与可解释性

DINOv2 Rocks Geological Image Analysis: Classification, Segmentation, and Interpretability

Akkodis Research(阿克奥迪斯研究院)

Florent Brondolo, Samuel Beaussant

AI总结:

本研究针对地球科学领域岩石 CT 图像分析数据稀缺、模型泛化差的问题,对比多种分割与分类方法,发现经 LoRA 微调的 DINOv2 在分布外任务中性能最优,分割结果甚至优于原始标注。

摘要:

计算机视觉领域的最新进展显著提升了图像分析任务的性能。然而,深度学习模型在应用于训练分布之外的领域时往往表现不佳,例如地球科学领域,该领域的特定领域数据可能较为稀缺。本研究针对岩石样本 CT 扫描图像的分类、分割与可解释性展开探究,重点关注现代计算机视觉技术在地球科学任务中的应用。我们对比了一系列分割方法,以评估它们在地质图像分析中的效能、效率与适配性。所评估的方法包括 Otsu 阈值法、聚类技术(K-means、模糊 C 均值)、一种监督机器学习方法(Random Forest)以及深度学习模型(UNet、ResNet152 和 DINOv2),实验采用了 10 个二分类砂岩数据集和 3 个多分类方解石数据集。DINOv2 因在特征提取方面表现出可观的成果,且具备应用于地球科学任务的潜力,被选中进一步评估其在处理 CT 扫描岩石数据时的可解释性与有效性。在分类任务中,未经微调的 DINOv2 在岩石图像分类中展现出强劲性能,即便 CT 扫描图像不在其原始训练集范围内亦是如此。在分割任务中,阈值法与聚类技术尽管计算效率较高,但即便经过预处理,所得结果仍不理想。相比之下,监督方法取得了更优的性能。深度学习方法虽需要更多计算资源,但所需人工干预极少,且具备更出色的泛化能力。尤其是经 LoRA 微调的 DINOv2,在分布外分割任务中表现优异,即便数据有限,在多分类任务中的表现也优于其他方法。值得注意的是,目视检查结果显示,DINOv2 生成的分割掩码往往比原始标注目标更准确。

英文摘要:

Recent advancements in computer vision have significantly improved image analysis tasks. Yet, deep learning models often struggle when applied to domains outside their training distribution, such as in geosciences, where domain-specific data can be scarce. This study investigates the classification, segmentation, and interpretability of CT-scan images of rock samples, focusing on the application of modern computer vision techniques to geoscientific tasks. We compare a range of segmentation methods to assess their efficacy, efficiency, and adaptability in geological image analysis. The methods evaluated include Otsu thresholding, clustering techniques (K-means, fuzzy C-means), a supervised machine learning approach (Random Forest), and deep learning models (UNet, ResNet152, and DINOv2), using ten binary sandstone datasets and three multi-class calcite datasets. DINOv2 was selected for its promising results in feature extraction and its potential applicability in geoscientific tasks, prompting further assessment of its interpretability and effectiveness in processing CT-scanned rock data. For classification, a non-fine-tuned DINOv2 demonstrates strong performance in classifying rock images, even when the CT-scans are outside its original training set. In segmentation tasks, thresholding and clustering techniques, though computationally efficient, produce subpar results despite preprocessing efforts. In contrast, supervised methods achieve better performance. While deep learning methods demand greater computational resources, they require minimal intervention and offer superior generalization. A LoRA fine-tuned DINOv2, in particular, excels in out-of-distribution segmentation and outperforms other methods in multi-class tasks, even with limited data. Notably, the segmentation masks generated by DINOv2 often appear more accurate than the original targets, based on visual inspection.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑