RAER
收藏资源简介:
RAER数据集是首个涵盖广泛自然学习环境的学术情感数据集,包含约2700个视频片段,来自大约140名学生的自然学习场景,如教室、图书馆、实验室和宿舍。数据集由两套学术情感标签进行标注,分别为粗粒度标签(专注或分心)和细粒度标签(享受、中性、困惑、疲劳或分心)。该数据集旨在帮助研究人员和教师准确识别学生在学习过程中的学术情感状态,从而调整教学策略,提高学习效果。
RAER dataset is the first academic emotion dataset that covers a broad spectrum of natural learning environments. It comprises approximately 2,700 video clips collected from natural learning scenarios of around 140 students, including classrooms, libraries, laboratories and dormitories. This dataset is annotated with two sets of academic emotion labels: coarse-grained labels (focused or distracted) and fine-grained labels (enjoyment, neutral, confused, fatigued, or distracted). The dataset is designed to assist researchers and teachers in accurately recognizing students' academic emotional states during the learning process, thereby adjusting teaching strategies and enhancing learning outcomes.
Context-Aware Academic Emotion Dataset and Benchmark (CAER)
基本信息
- 会议: ICCV 2025
- 作者: Luming Zhao1*, Jingwen Xuan1*, Jiamin Lou2, Yonghui Yu1, Wenwu Yang1†
- 机构: 1浙江工商大学, 2浙江越秀外国语学院
- 贡献: *Equal Contribution, †Corresponding Author
数据集概述
- 名称: RAER (Context-Aware Academic Emotion Dataset)
- 内容: 约2,700个视频片段
- 采集场景: 自然学习环境(教室、图书馆、实验室、宿舍等)
- 参与者: 约140名学生
- 标注: 每个视频片段由约10名标注者独立标注,使用两种不同粒度的学术情感标签
研究背景
- 目标: 通过面部表情自动识别真实学习环境中的学术情感
- 挑战: 学术情感识别领域缺乏公开数据集
方法创新
- 提出方法: CLIP-CAER (CLIP-based Context-aware Academic Emotion Recognition)
- 特点: 利用可学习文本提示整合面部表情和上下文信息
- 优势: 显著优于现有基于视频的面部表情识别方法
引用格式
bibtex @InProceedings{Zhao_2025_ICCV, author = {Zhao, Luming and Xuan, Jingwen and Lou, Jiamin and Yu, Yonghui and Yang, Wenwu}, title = {Context-Aware Academic Emotion Dataset and Benchmark}, booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)}, year = {2025} }
致谢
感谢所有为本研究提供支持的老师和学生




