MidiCaps

该数据集名为MidiCaps，包含了文本描述和MIDI文件的对，用于评估符号音乐生成模型的有效性。在研究中，该数据集既用于主观评估，也用于客观评估。主观评估包括听力研究，而客观评估则通过测量与音乐生成相关的多种指标来进行。尽管具体数量未提供，但该数据集包含了多个用于评估的示例，其任务专注于符号音乐生成。

This dataset, named MidiCaps, consists of pairs of text descriptions and MIDI files, and is used to evaluate the effectiveness of symbolic music generation models. In relevant research, this dataset is applied for both subjective and objective evaluations. Subjective evaluations include listening studies, while objective evaluations are conducted by measuring multiple metrics related to music generation. Although the exact number of samples is not provided, the dataset contains multiple evaluation examples, with its tasks focused on symbolic music generation.