EUVIMEDEUVIMED entdecken
Crossref

Hybrid Transformer-CNN Architecture for Efficient Point Cloud Classification with Attention Mechanisms

Chiyu Wang

Innovative Applications of AI · 2024

DOI: 10.70695/5ck0y871 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Innovative Applications of AI, 2024. Verfasst von Chiyu Wang.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This paper proposes a Transformer-based point cloud classification method. By designing two key modules: a hybrid feature extraction module and a multi-head self-attention module, efficient feature extraction and classification of point cloud data are achieved. The hybrid feature extraction module combines the advantages of convolutional neural network (CNN) and Transformer architecture, extracts local geometric features through convolution operations. The multi-head self-attention module captures different feature relationships in the point cloud through multiple self-attention heads to further enhance feature representation. Experimental results also demonstrate the effectiveness of our method. Our method achieves 91.6% mAcc and 93.3% OA on ModelNet40 dataset; on the ScanObjectNN dataset, it achieves 79.8% mAcc.

Zitieren

APA

Wang, C. (2024). Hybrid Transformer-CNN Architecture for Efficient Point Cloud Classification with Attention Mechanisms. Innovative Applications of AI. https://doi.org/10.70695/5ck0y871

Vancouver

Wang C. Hybrid Transformer-CNN Architecture for Efficient Point Cloud Classification with Attention Mechanisms. Innovative Applications of AI. 2024. doi:10.70695/5ck0y871

BibTeX
@article{Wang2024,
  author = {Wang, Chiyu},
  title = {Hybrid Transformer-CNN Architecture for Efficient Point Cloud Classification with Attention Mechanisms},
  journal = {Innovative Applications of AI},
  year = {2024},
  doi = {10.70695/5ck0y871}
}
Kostenlos starten →