# Micro-Expression Recognition via LoRA-Enhanced DinoV2 and Interactive Spatio-Temporal Modeling

## TL;DR

The paper introduces a new method for recognizing micro-expressions using advanced machine learning techniques that improve accuracy and reduce computational costs.

## Contribution

A novel MER architecture combining LoRA, frequency-domain transformation, and graph-based temporal modeling is introduced.

## Key findings

- The proposed method achieves an 81.16% UF1 score on the SAMM dataset.
- It outperforms existing methods by 0.96% in UF1 and 2.27% in UAR on the SAMM dataset.
- The DGAT network effectively models temporal features using graph-based attention.

## Abstract

Micro-expression recognition (MER) is challenged by a brief duration, low intensity, and heterogeneous spatial frequency patterns. This study introduces a novel MER architecture that reduces computational cost by fine-tuning a large feature extraction model with LoRA, while integrating frequency-domain transformation and graph-based temporal modeling to minimize preprocessing requirements. A Spatial Frequency Adaptive (SFA) module decomposes high- and low-frequency information with dynamic weighting to enhance sensitivity to subtle facial texture variations. A Dynamic Graph Attention Temporal (DGAT) network models video frames as a graph, combining Graph Attention Networks and LSTM with frequency-guided attention for temporal feature fusion. Experiments on the SAMM, CASME II, and SMIC datasets demonstrate superior performance over existing methods. On the SAMM 5-class setting, the proposed approach achieves an unweighted F1 score (UF1) of 81.16% and an unweighted average recall (UAR) of 85.37%, outperforming the next best method by 0.96% and 2.27%, respectively.