Screw Tightening Action Detection Dataset

In the industrial field, with the increase in automation, the detection of tightening actions has become particularly important. Currently, many enterprises face the problem of identifying and monitoring the tightening process by operators and robots. Traditional methods often rely on manual inspection, which is inefficient and prone to error. By establishing a screw tightening action detection dataset, the aim is to provide training data for smart manufacturing, enhancing the accuracy of automatic detection systems. This dataset consists of multiple videos covering various tightening actions, effectively supporting the training and validation of machine learning models. Data collection uses high-resolution cameras in a standardized factory environment to ensure video quality. For quality control, we adopt a multi-round annotation and expert review approach to ensure annotation consistency and accuracy. Data is stored in MP4 format for easy subsequent processing and analysis.