SULAND v2: A Refined RGB Dataset and Deep Learning Object Detection Benchmark for UAV/UGV-Based SUrface LANDmine Detection Under Domain Shift
Quick Answer
SULAND_v2 introduces a refined RGB dataset for UAV/UGV-based landmine detection, featuring 33,771 images and 12,433 bounding boxes.
Quick Take
Enhanced annotations improve YOLOv8's in-distribution mAP@50 by 14.6-19.6 points, while correcting the OOD class-ID convention boosts OOD performance by ~25 points. YOLOv12-Small achieves the highest IID mAP@50 of 0.908, underscoring the importance of robust benchmarks in safety-critical applications.
Key Points
- SULAND_v2 contains 33,771 images and 12,433 bounding boxes for landmine detection.
- Annotation refinements improve YOLOv8's IID mAP@50 by 14.6-19.6 percentage points.
- Correcting the OOD class-ID convention increases YOLOv8's OOD mAP@50 by ~25 points.
- YOLOv12-Small achieves the highest IID mAP@50 of 0.908 on SULAND_v2.
- The benchmark highlights the gap between high IID accuracy and operational readiness.
DeepSignal Analysis
What happened
SULAND_v2 is a refined RGB dataset aimed at improving UAV/UGV-based landmine detection. It includes 33,771 images and 12,433 bounding boxes, with improved annotations that enhance YOLOv8's performance significantly. The dataset addresses previous issues in SULAND, such as annotation errors and class-ID conventions.
Key evidence
- SULAND_v2 contains 33,771 images and 12,433 bounding boxes, providing a substantial dataset for landmine detection.
- Annotation refinements improved YOLOv8's in-distribution mAP@50 by 14.6-19.6 percentage points, indicating enhanced detection capabilities.
- YOLOv12-Small achieved the highest IID mAP@50 of 0.908 on SULAND_v2, highlighting the dataset's effectiveness in benchmarking.
Why it matters
The development of SULAND_v2 is crucial for advancing object detection in safety-critical applications like landmine detection. By addressing previous dataset shortcomings, it provides a more reliable benchmark for evaluating the robustness of detection algorithms under varying conditions. This is particularly important given the limited availability of public datasets in this domain.
What to watch
Paper Resources
📖 Reader Mode
~2 min readAbstract:RGB imagery offers a practical, low-cost option for Unmanned Aerial/Ground Vehicle (UAV/UGV) survey support in surface-landmine detection, but object detectors remain underexplored in this safety-critical domain. Limited cross-architecture benchmarking and insufficient out-of-distribution (OOD) analysis obscure whether detectors generalize across deployment conditions. This challenge is amplified by the scarcity of public RGB landmine datasets, making SULAND a key benchmark for PFM-1 and PMA-2 detection. However, inspection reveals missing/false annotations, localization errors, inconsistent visibility criteria, visual artifacts, temporal labeling inconsistencies, and an inverted OOD class-ID convention in SULAND. We present SULAND_v2, a refined RGB surface-landmine dataset and benchmark. Preserving original images and splits, we manually revise annotations to ensure completeness, precise localization, label validity, and class consistency. SULAND_v2 contains 33,771 images and 12,433 bounding boxes. We benchmark 35 detector configurations across nine families. Annotation refinement improves YOLOv8 in-distribution (IID) test mAP@50 by 14.6-19.6 percentage points, while fixing the OOD class-ID convention increases mean YOLOv8 OOD mAP@50 by ~25 percentage points. On SULAND_v2, YOLOv12-Small achieves the highest IID mAP@50 (0.908), while RF-DETR-Large yields the strongest OOD performance (0.799 mAP@50, 0.675 recall). Our results demonstrate that high IID accuracy does not guarantee operational readiness. SULAND_v2 provides a reliable benchmark for evaluating domain-shift robustness in RGB-based mine-action survey support.
| Comments: | The manuscript is currently under submission to a journal for peer review |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2607.28996 [cs.CV] |
| (or arXiv:2607.28996v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2607.28996 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Sagar Lekhak [view email]
[v1]
Fri, 31 Jul 2026 03:51:17 UTC (20,839 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →ProMoE-FL: Prototype-conditioned Mixture of Experts for Multimodal Federated Learning with Missing Modalities
ProMoE-FL introduces a Prototype-conditioned Mixture-of-Experts framework for multimodal federated learning, effectively addressing missing modalities. It outperforms existing methods on four chest X-ray datasets, demonstrating superior feature synthesis capabilities in both homogeneous and heterogeneous settings.