Cotton Module Cover Damage Dataset (CMCD)

The CMCD dataset contains RGB images of round cotton modules with damaged and undamaged plastic covers, captured in gin yards and during real-world handling operations. Training images were collected using cell phone cameras while the modules were staged in the gin yards and were augmented to 7,200 images with 10,068 bounding box annotations for CNN training and validation. An independent test set of 250 images provides image-level labels for evaluation. The real-world dataset includes images of cotton modules captured during handling with loaders (22,736 images) and yard trucks (34,921 images) using integrated imaging systems, used to evaluate the trained model’s ability to detect damaged covers under operational conditions. This dataset is intended for research on cover damage detection in round cotton modules, aiming to reduce plastic contamination and preserve cotton quality (https://doi.org/10.1016/j.compag.2025.111023). This dataset could be helpful for studies on autonomous cotton module handling, logistics, and machine-vision applications in real-world scenarios.