Abstract
We present a facial emotion recognition framework, built upon Swin vision Transformers jointly with squeeze and excitation block (SE). A transformer model based on an attention mechanism has been presented recently to address vision tasks. Our method uses a vision transformer with a Squeeze excitation block (SE) and sharpness-aware minimizer (SAM). We have used a hybrid dataset, to train our model and the AffectNet dataset to evaluate the result of our model
Abstract (translated)
我们提出了一种面部情感识别框架,该框架基于 Swin 视觉Transformer 和挤压和激励块(SE)构建而成。最近,提出了一种基于注意力机制的Transformer模型,以解决视觉任务。我们的方法使用了一个视觉Transformer,带有挤压激励块(SE)和尖化 aware 最小化器(SAM)。我们使用了混合数据集,用于训练我们的模型和AffectNet数据集来评估我们的模型的结果。
URL
https://arxiv.org/abs/2301.10906
