SULAND v2: A Refined RGB Dataset and Deep Learning Object Detection Benchmark for UAV/UGV-Based Surface Landmine Detection Under Domain Shift

Researchers released SULAND_v2, a corrected RGB dataset for surface landmine detection.

They manually fixed annotation errors and class-consistency issues in the original SULAND while keeping the same images and splits.

The updated set includes 33,771 images and 12,433 bounding boxes, and benchmarks 35 detection configurations across nine model families.

Results show strong in-distribution performance but major drops under domain shift and out-of-distribution handling.

This benchmark offers a more reliable test of generalization for safety-critical UAV and UGV mine-action survey support.