Audio-Visual Cutting Event Dataset

To the best of our knowledge, no existing public dataset adequately meets the requirements of our specific task. Consequently, we developed the Audio-Visual Cutting Event (AVCE) dataset tailored to coal mining scenarios. This dataset comprises 120.35 hours of footage obtained from 16 cameras strategically placed within these environments. The video data were segmented into 10-second intervals, resulting in a total of 1796 clips. Each segment was meticulously annotated to indicate the occurrence of specific events, which were subsequently categorized into seven distinct groups. Notably, every video clip encompasses at least 2 seconds of audio-visual data relevant to these events, with a substantial 84.6% of clips containing full 10-second events. The dataset encapsulates a variety of event types common in coal mining operations, such as coal and rock cutting, as well as the idling and stopping of shearers.