STRIX_3_dataset_merged
收藏资源简介:
mad-esc50-drone-audio 是一个用于音频分类任务的融合数据集,它整合并标准化了来自三个知名开源音频数据源的内容:MAD (Military Audio Dataset)、ESC-50 和 DroneAudioDataset。该数据集旨在提供一个多样化的音频样本集合,特别包含了军事音频、环境声音和无人机音频等类别。数据集经过处理,以避免源数据之间的重复,例如移除了 DroneAudioDataset 中源自 ESC-50 的 unknown 子集。数据以结构化形式组织,包含四个关键字段:音频数据 (`audio`)、类别标签 (`label`)、数据来源 (`source_dataset`) 以及数据划分 (`split`)。数据规模属于 1B<n<10B 类别。数据集共定义了 59 个细粒度的音频类别,涵盖从数字标签(0-6)到丰富的具体声音,如飞机、汽车喇叭、猫叫、链锯、鸟鸣、教堂钟声、咳嗽、狗吠、脚步声、直升机、笑声、键盘敲击、雨声、海浪、警笛、打喷嚏、雷暴、火车、吸尘器等广泛的环境、动物、机械和人类活动声音。该数据集适用于音频分类模型的研究、训练与评估,尤其关注军事、环境和无人机相关的声音识别场景。其许可证为 CC-BY-NC-3.0,限制非商业用途。
mad-esc50-drone-audio is a fused dataset for audio classification tasks. It integrates and standardizes content from three prominent open-source audio datasets: MAD (Military Audio Dataset), ESC-50, and DroneAudioDataset. This dataset aims to provide a diverse collection of audio samples, specifically encompassing categories such as military audio, environmental sounds, and drone audio. The dataset has been processed to eliminate duplication across source datasets, for example, removing the unknown subset from DroneAudioDataset that originates from ESC-50. The data is organized in a structured format with four key fields: audio data (`audio`), category label (`label`), data source (`source_dataset`), and data split (`split`). The dataset size falls within the range of 1B < n < 10B. A total of 59 fine-grained audio categories are defined in this dataset, covering a broad spectrum of environmental, animal, mechanical, and human activity sounds, ranging from numeric labels (0-6) to a wide array of specific sounds including airplanes, car horns, cat meows, chainsaws, bird chirps, church bells, coughs, dog barks, footsteps, helicopters, laughter, keyboard taps, rain sounds, ocean waves, sirens, sneezes, thunderstorms, trains, and vacuum cleaners. This dataset is applicable to the research, training, and evaluation of audio classification models, with a particular focus on sound recognition scenarios related to military, environmental, and drone applications. Its license is CC-BY-NC-3.0, which restricts non-commercial usage.
数据集概述
该数据集是一个融合音频分类数据集,由三个来源的音频数据合并而成,经过标准化处理后包含 audio、label、source_dataset、split 四个字段。
数据集名称:Fusion MAD + ESC-50 + DroneAudioDataset (mad-esc50-drone-audio)
许可证:CC BY-NC 3.0
标签:音频、音频分类、军事音频、无人机检测、环境声音
数据规模:1B < n < 10B
数据来源
- MAD(Military Audio Dataset):Kim, Yoon & Jung, 2024,发布于 Scientific Data,采用 CC BY 4.0 许可证。来源:https://www.kaggle.com/datasets/junewookim/mad-dataset-military-audio-dataset
- ESC-50:Karol J. Piczak,采用 CC BY-NC 3.0 许可证(仅限非商业使用)。来源:https://github.com/karolpiczak/ESC-50
- DroneAudioDataset:Al-Emadi 等,IWCMC 2019,学术使用需引用论文。来源:https://github.com/saraalemadi/DroneAudioDataset
注意:DroneAudioDataset 中的 "unknown" 子集因复用了 ESC-50 文件,为避免重复已被排除在融合之外(详见构建 notebook)。
类别信息
数据集共包含 59 个类别,涵盖通用环境声音、军事音频与无人机音频。部分类别列表如下:
0、1、2、3、4、5、6、7、airplane、bebop_1、breathing、brushing_teeth、can_opening、car_horn、cat、chainsaw、chirping_birds、church_bells、clapping、clock_alarm、clock_tick、coughing、cow、crackling_fire、crickets、crow、crying_baby、dog、door_wood_creaks、door_wood_knock、drinking_sipping、engine、fireworks、footsteps、frog、glass_breaking、hand_saw、helicopter、hen、insects、keyboard_typing、laughing、membo_1、mouse_click、pig、pouring_water、rain、rooster、sea_waves、sheep、siren、sneezing、snoring、thunderstorm、toilet_flush、train、vacuum_cleaner、washing_machine、water_drops、wind




