An open paradigm dataset for intelligent monitoring of underground drilling operations in coal mines

In the field of coal mine safety monitoring and automation, high-quality and specialized datasets are crucial for the development and validation of artificial intelligence algorithms. Currently, there is no comprehensive benchmark dataset specifically for coal mine industrial scenarios, which significantly limits the research progress of AI algorithms in the coal mining industry. This study has constructed for the first time a benchmark dataset (DsDPM 66) specifically for coal mine heading faces, containing 105,096 images obtained from videos of 66 drilling operation scenes. The dataset has been meticulously annotated manually to suit computer vision tasks such as object detection and pose estimation. In addition, this study conducted extensive benchmarking experiments on this dataset, applying various advanced AI algorithms including but not limited to YOLOv8 and DETR. The experimental results show that the proposed dataset can effectively improve the accuracy of various object detection and pose estimation models in coal mines, filling the data gap in the coal mining field and providing valuable resources for the development of coal mine safety monitoring and automation technologies.