M4-Instruct-Data

M4-Instruct数据集专注于多模态模型的训练，特别是针对交错多图像处理能力。该数据集包含从公共数据集收集或由GPT-4V API生成的多图像数据，包括多帧视频和多视角3D数据，规模在10万到100万之间。数据集以json文件形式提供标注信息，并提供配套图像数据。遵循Creative Commons Attribution 4.0 International许可协议，并遵守OpenAI的使用政策。主要用于支持视觉问答和问答任务的研究。

The M4-Instruct dataset focuses on the training of multimodal models, especially for interleaved multi-image processing capabilities. It contains multi-image data collected from public datasets or generated via the GPT-4V API, including multi-frame videos and multi-view 3D data, with a scale ranging from 100,000 to 1,000,000 entries. The dataset provides annotation information in JSON file format, along with supporting image data. It adheres to the Creative Commons Attribution 4.0 International License and complies with OpenAI's usage policies. It is primarily used to support research on visual question answering (VQA) and question answering (QA) tasks.