End-to-End Microexpression Detection Using 3D Convolution and LSTM

Hosen M. I., Gülcü A.

14th International Conference on Image Processing, Theory, Tools and Applications, IPTA 2025, İstanbul, Türkiye, 13 - 16 Ekim 2025, (Tam Metin Bildiri) identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/ipta66025.2025.11222042
  • Basıldığı Şehir: İstanbul
  • Basıldığı Ülke: Türkiye
  • Anahtar Kelimeler: 3D Convolution, Deep Learning, LSTM, Microexpression
  • İstanbul Ticaret Üniversitesi Adresli: Evet

Özet

Micro-expressions are brief facial expressions that last only a fraction of a second. Micro-expressions is challenging due to subtle spatiotemporal dynamics. Existing methods often requires additional steps such as apex spoting and optical flow calculation that hinder end-to-end learning. In this work, we propose a novel end-to-end deep learning framework that integrates a 3D Convolutional Neural Network with Long Short-Term Memory (LSTM) networks to effectively capture micro-expressions. Our architecture employs a narrow deeper 3D convolutional backbone followed by an LSTM layer to enhance temporal modeling while maintaining computational efficiency. Extensive experiments on the CASME II dataset demonstrate that the proposed model achieves a overall accuracy of 72.11 %, significantly outperforming baseline architecture.