What question did this study set out to answer?

To develop an improved U-Net architecture for accurate crop and weed segmentation from aerial images.

May 13, 2026SensorsOpen Access

A U-Net Improved Version for Crop and Weed Segmentation from Aerial Images

View Full Paper

Ask AI

Bookmark

Authors

Novel U-Net architecture improves segmentation accuracy in crops and weeds, indicating better efficiency for Precision Agriculture.

To develop an improved U-Net architecture for accurate crop and weed segmentation from aerial images.
Proposed U-Net variant includes deep supervision and a Vegetation Global Context block.
Dual-headed output model predicts separate masks for crops and weeds.
Evaluation performed using multiple encoder backbones: ConvNextV2, FastViT, RepViT, and MambaVision.
Achieved a maximum of 94.24% Dice score for crop segmentation and 93.72% for weeds.
Reduced inference time by eliminating multi-stage processing, suitable for real-time deployment.

Bunica-Mihai et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark