InCarEmo
收藏资源简介:
InCarEmo是由哈尔滨工业大学与赛力斯联合创建的多模态车载情感识别与驾驶员状态监测数据集,旨在填补现有车载情感计算数据在模态多样性和对话语境方面的空白。该数据集包含6,100条样本,涵盖7.5小时的多模态记录,集成了RGB与红外视频、车内音频及对话文本,数据来源于脚本化但贴近真实的车载交互场景,并额外构建了英文基准以支持跨语言评估。数据集通过精心设计的采集流程构建,包括脚本生成、视频采集与专家标注,确保了数据质量与标注一致性。其核心应用领域为智能座舱系统,致力于解决驾驶员情感识别、疲劳检测与分心监测等关键问题,以提升驾驶安全与人车交互体验。
ClassEval-Pro is a cross-domain class-level code generation benchmark dataset jointly constructed by Shanghai Jiao Tong University and other institutions. It includes 300 tasks covering 11 domains. This dataset is built via an automated three-stage pipeline, integrating real GitHub code from after 2025, with an average of 113 lines of code per task, and validated by strict 90% line coverage tests. The dataset aims to evaluate the cross-method coordination capability of large language models (LLMs) when constructing complete class structures, addressing the shortcomings of existing benchmarks in the evaluation of compositional code creation.



