MUImage

MUImage Dataset

This is the MUImage dataset used to facilitate image to music generation, consisting of 9,966 music files with a total playtime of 27.72 hours generated using the MU-LLaMA and BLIP Image Captioning models.
This dataset is used to train the M2UGen model.
The MUImageInstructions.json file contains a list with each of the element having the following format:
{
"input_file": "1OeqDC6vmto.jpg",
"output_file": "1OeqDC6vmto.mp3",
"conversation": [
{… See the full description on the dataset page: https://huggingface.co/datasets/M2UGen/MUImage.

MUImage 数据集系为图像至音乐生成而设计的专用数据集，包含由 MU-LLaMA 和 BLIP 图像标题模型生成的 9,966 个音乐文件，总时长达 27.72 小时。该数据集旨在用于训练 M2UGen 模型。MUImageInstructions. 文件列出了每个元素，其格式如下：
{
"input_file": "1OeqDC6vmto.jpg",
"output_file": "1OeqDC6vmto.mp3",
"conversation": […]
}
详细描述请参阅数据集页面：https://huggingface.co/datasets/M2UGen/MUImage。