Image Recognition of Textile Healing Effect Based on Transformer-LSTM
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
To address the challenge of quantifying cross-modal correlations between textile microstructure and dynamic physiological responses in therapeutic efficacy evaluation, this paper proposes a hybrid model integrating standard Transformer and LSTM. The dual-stream architecture synergistically analyzes static image features and temporal biometric feedback: the main network employs a standard Transformer encoder to extract surface features such as micrometer-scale fiber orientation and pore distribution, while the LSTM branch processes physiological sequences collected during wear. A cross-attention mechanism establishes quantitative mappings between textile structural parameters and physiological indicators. Validated on 12,480 multimodal datasets, the model achieves an F1-score of 0.927 in oven cotton classification and the average F1-score reaches 0.901. This work demonstrates the effectiveness of standard Transformer-LSTM in fusing static and temporal data, providing a scalable solution for intelligent textile efficacy assessment.
Fields of Study & Subjects
Multimodal Machine Learning Applications
Textile materials and evaluations
Advanced Neural Network Applications
Publication details
- Journal / source
- IEEE International Conference on Power, Electronics and Green Energy (ICPEGE)
- Pages
- 336–339
- Publication date
- 28 January 2026
- Language
- EN