中文

第 6 届 ABAW 挑战赛中基于视觉语言模型的零样本复合表情识别

计算机视觉与模式识别 2024-03-19 v1

摘要

传统的面部表情识别方法主要集中于六种基本面部表情的分类。然而,由于全面的训练数据集有限,现实世界场景呈现出更广泛的复杂复合表情,这些表情由基本表情组合而成。第 6 届野外情感行为分析研讨会与竞赛 (ABAW) 提供了包含复合表情的未标注数据集。在本研究中,我们提出了一种零样本方法,通过利用预训练的视觉语言模型并结合一些传统的 CNN 网络来识别复合表情。

关键词

引用

@article{arxiv.2403.11450,
  title  = {Zero-shot Compound Expression Recognition with Visual Language Model at the 6th ABAW Challenge},
  author = {Jiahe Wang and Jiale Huang and Bingzhao Cai and Yifan Cao and Xin Yun and Shangfei Wang},
  journal= {arXiv preprint arXiv:2403.11450},
  year   = {2024}
}

备注

USTC-AC's paper for Compound Expression (CE) Recognition Challenge in 6th Workshop and Competition on Affective Behavior Analysis in-the-wild (ABAW)