Optimizing precision agriculture: A real-time detection approach for grape vineyard unhealthy leaves using deep learning improved YOLOv7 with feature extraction capabilities

Zohaib Khan, Hui Liu, Yue Shen, Zhaofeng Yang, Lanke Zhang, Feng Yang

Abstract: Highlights•Introduced Light Convolution into YOLOv7 backbone achieving mAP 64.6 %, P 73.7 %, R 65.5 %.•Improvement of results in Precision 3.2 %, Recall 6.2 %, mean Average Precision 1.6 %.•Improvement of 65.96 % of spraying pesticide on unhealthy leaves of grape vineyard.•GELU and ELU activation function used with Squeeze and Excitation Block.•Adaptive gradient optimizer used with lasso regularization technique.