Scholarlink

<p>Minerals have always played a pivotal role in human progress, from ancient societies to contemporary technological and industrial uses. Their global accessibility and strategic significance spans from essential raw materials in manufacturing to crucial components of nuclear power. To explore and find a solution to this issue, the given paper introduces a convolution-fused mobile friendly vision transformer (MobileViT-XS) for multi-class minerals classification. The study introduces a hybrid deep learning-based mineral classification framework optimized for mobile applications, leveraging convolutional feature extraction to capture subtle visual characteristics of mineral samples. MobileViT-XS enables the identification of the minerals from laboratories or fields on mobile phones or edge devices at an ease. The results demonstrate that the proposed framework exhibits superior feature extraction capabilities, with MobileViT-XS achieving an accuracy of 92.75% thus outperforming conventional transformer models such as data efficient image transformer (DeiT), MobileNetV3, ShuffleNetV2 and many more. Overall, the given study contributes to the growing body of research in mobile computer vision and lays the groundwork for intelligent, edge-deployable mineral identification systems in geoscientific and industrial appliances.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Conv_MobileViT-XS: hybrid convolutional fused mobile friendly vision transformer framework for classification of minerals

  • 期刊论文

  • 发表日期：

    2025年12月8日

查看全文
  • Shally Vats,
  • Shuchita Koul,
  • Vijay Shankar Sharma

摘要

Minerals have always played a pivotal role in human progress, from ancient societies to contemporary technological and industrial uses. Their global accessibility and strategic significance spans from essential raw materials in manufacturing to crucial components of nuclear power. To explore and find a solution to this issue, the given paper introduces a convolution-fused mobile friendly vision transformer (MobileViT-XS) for multi-class minerals classification. The study introduces a hybrid deep learning-based mineral classification framework optimized for mobile applications, leveraging convolutional feature extraction to capture subtle visual characteristics of mineral samples. MobileViT-XS enables the identification of the minerals from laboratories or fields on mobile phones or edge devices at an ease. The results demonstrate that the proposed framework exhibits superior feature extraction capabilities, with MobileViT-XS achieving an accuracy of 92.75% thus outperforming conventional transformer models such as data efficient image transformer (DeiT), MobileNetV3, ShuffleNetV2 and many more. Overall, the given study contributes to the growing body of research in mobile computer vision and lays the groundwork for intelligent, edge-deployable mineral identification systems in geoscientific and industrial appliances.