A tiny object detection dataset for high-altitude open-water search and rescue
Organizations: Hangzhou International Innovation Institute of Beihang University, Beihang University, Hangzhou 311100, China · School of Transportation Science and Engineering, Beihang University, Beijing 100191, China · GRG Banking Equipment Co., Ltd.
Abstract
Large aerial platforms often operate above 1.5 km to improve the efficiency of open water search and rescue (OW-SAR). However, long imaging distances, wide fields of view, and limited sensor resolution reduce objects to a few pixels, increasing their susceptibility to environmental interference and making detection harder than for conventional small objects. To our knowledge, existing datasets focus on low-altitude scenarios and do not support automatic detection at 1.5-2.0 km. We therefore present TinyWOD, a tiny object detection dataset for high-altitude OW-SAR using lifebuoys as detection targets. It comprises 11,359 manually annotated images and 19,505 annotated instances spanning diverse environments, lighting conditions, water colors, and surface ripples. Compared with existing datasets, TinyWOD contains much smaller objects and a substantially higher proportion of tiny instances; its average equivalent side length (ESL) is 12.7 pixels, with 76.96% of in stances having ESLs below 15 pixels, reflecting actual high-altitude water-rescue imaging conditions. We define APtiny, a tiny object detection metric, and benchmark mainstream and state-of-the-art networks using this and conventional metrics. All evaluated networks effectively detect tiny waterborne objects, demonstrating TinyWOD's reusability and utility. We expect this dataset to support research on tiny object detection in high-altitude OW-SAR.
Figures & tables
| Basic Information | Box Description | Environmental Factors | |||||||||||
| Dataset | # Images | Res. (main) | Avg. box area (px 2 ) | Average box-to-image area ratio (%) | Average ESL | ESL 15 (%) | Illumination | Water Color | Surface | ||||
| Day | Night | Blue | Green | Yellow | Calm | Rough | |||||||
| SeaDronesSee [ 18 ] | 14,227 | 9004.5 | 0.130 | 62.6 | 3.50 | ||||||||
| MOBDrone [ 19 ] | 126,170 | 6506.1 | 0.335 | 60.4 | 1.71 | ||||||||
| AFO [ 17 ] | 3,647 | 10392.5 | 0.147 | 91.3 | 0.03 | ||||||||
| SeaPerson [ 20 ] | 12,032 | 661.7 | 0.031 | 22.7 | 26.68 | ||||||||
| Method | Params (M) | GFLOPs | FPS | Accuracy | Precision | Recall | AR 100 | ||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| AP | AP 50 | AP 75 | AP small | ||||||||
| YOLO26s [ 34 ] | 9.95 | 22.5 | 407.3 | 0.404 | 0.689 | 0.436 | 0.403 | 0.320 | 0.650 | 0.558 | 0.492 |
| YOLO26x [ 34 ] | 55.64 | 209.1 | 220.9 | 0.443 | 0.723 | 0.494 | 0.442 | 0.367 | 0.789 | 0.595 | 0.513 |
| VMINet-B + Faster R-CNN [ 35 , 29 ] | 49.12 | 97.3 | 42.3 | 0.282 | 0.587 | 0.252 | 0.281 | 0.171 | 0.697 | 0.538 | 0.351 |
| DINO-R50 [ 36 ] | 46.82 | 983.9 | 45.1 | 0.386 | 0.713 | 0.368 | 0.385 | 0.300 | 0.476 | 0.416 | 0.501 |
| H-DETR-R50 [ 37 ] | 47.78 | 343.4 | 25.0 | 0.431 | 0.788 | 0.418 | 0.430 | 0.351 | 0.787 | 0.669 | 0.553 |
Aerial Object Detection
Aerial object detection locates and classifies objects in drone and satellite imagery, often under scale, orientation, clutter, and compute constraints.
See all 132 papers