MouseGPT Open-Vocabulary Dataset
收藏资源简介:
MouseGPT开放词汇数据集是由上海科技大学信息科学与技术学院和生命科学与技术学院创建的大型多模态数据集。该数据集包含超过4200万帧多样精神状况下小鼠的多视角视频记录,如抑郁、幻觉和精神分裂症等。数据集通过精确的3D关键点位置捕捉,确保了对自然和刺激诱发行为的详细记录。该数据集为MouseGPT模型的监督微调提供了多样化的基础,使模型能够生成对小鼠行为的详细、上下文敏感的描述,从而为神经精神疾病研究提供了深入了解复杂行为的灵活方法。
The MouseGPT Open-Vocabulary Dataset is a large-scale multimodal dataset developed by the School of Information Science and Technology and the School of Life Science and Technology at ShanghaiTech University. This dataset encompasses over 42 million frames of multi-view video recordings of mice across diverse mental states, such as depression, hallucinations, schizophrenia and other related conditions. Precise 3D keypoint localization is employed to ensure detailed documentation of both spontaneous and stimulus-induced behaviors. This dataset serves as a diversified foundational resource for the supervised fine-tuning of the MouseGPT model, enabling the model to generate detailed, context-aware descriptions of mouse behaviors, thereby providing a flexible methodology for gaining in-depth insights into complex behaviors in neuropsychiatric disease research.




