IndianCropResidue: A pioneering dataset and its efficient segmentation for optimal crop residue detection and management

Open‐field burning of crop residues in India remains a major contributor to air pollution and soil degradation, largely driven by economic constraints and the limited turnaround window between harvesting and sowing of the succeeding crop. Although sustainable alternatives such as unmanned aerial vehicle (UAV)‐assisted spraying of bio‐decomposers have been proposed, their effectiveness is hindered by the lack of high‐quality annotated datasets for accurately mapping uneven residue distribution. To bridge this gap, this study introduces IndianCropResidue, a structured dataset designed for deep learning‐based crop‐residue monitoring under real field conditions. An attention‐enhanced ResNet‐50‐driven convolutional neural network incorporating channel–spatial attention and multiscale pooling delivers precise segmentation and density estimation that enables targeted bio‐decomposer application by UAVs. The proposed framework achieved an overall accuracy of 92.5% and an F1‐score of 92.3%, demonstrating consistent segmentation performance across heterogeneous field conditions. The results highlight the potential of data‐driven residue mapping to support targeted UAV‐based bio‐decomposer applications and improved precision residue management.