Frame-to-Panorama Localization and Context-Aware Sampling for Scene-Specific Ship Detection in a Smart Marina Testbed

Ignat Romanov; Andreas Hadjipieris; Neofytos Dimitriou

cs.CVSep 24, 2026

Frame-to-Panorama Localization and Context-Aware Sampling for Scene-Specific Ship Detection in a Smart Marina Testbed

Authors: Ignat Romanov, Andreas Hadjipieris, Neofytos Dimitriou

Organizations: Department of Computer Science University of Nicosia Nicosia, Cyprus · Maritime Digitalization Centre Cyprus Marine and Maritime Institute Larnaca, Cyprus

Abstract

Smart maritime infrastructures provide continuous access to heterogeneous sensing streams, enabling repeated experimentation, digital-twin development, and AI-based maritime services. However, sensing hardware alone is not sufficient for scene-specific model development: historical video streams must also be spatially indexed, contextualized, and reduced to informative subsets for annotation. This paper presents a frame-to-panorama localization and context-aware sampling pipeline for ship detection in historical PTZ maritime video lacking reliable pan, tilt, and zoom metadata. The main contribution is an end-to-end data-curation approach that recovers camera-view information from historical PTZ video and combines it with environmental context and visual diversity to construct compact, scene-specific training sets. Specifically, frames are localized on a reference panorama using SuperPoint and LightGlue, enriched with weather and solar-state metadata, and selected through diversity sampling to preserve variation across camera view and environmental conditions. A second context-aware stage targets under-represented distant-vessel cases near the horizon using tile-level visual embeddings and Gaussian Mixture Model clustering. Applied within the CMMI MDigi-I Smart Marina testbed, the proposed pipeline reduces 40,718 candidate frames to 220 images for annotation, corresponding to a 99.5% reduction. A YOLO26-m detector fine-tuned on this subset achieves a mean AP50 of 94.78% $\pm$ 0.51% and a mean AP50-95 of 75.10% $\pm$ 1.73% under sequence-grouped five-fold cross-validation. These results demonstrate that highly redundant infrastructure video streams can be transformed into compact, spatially and contextually diverse training sets for scene-specific detector adaptation while substantially reducing annotation effort.

Figures & tables

Fig. 1: Example of a video frame on the left. The center of this example is mapped to the panorama perspective through a homography transformation as shown on the right.

Fig. 2: Spatial distribution of mapped video frames on the reference panorama. The heatmap highlights frequently observed camera poses, including the dominant default park position.

Algorithm	Accuracy	Macro Average			Weighted Average
Algorithm	Accuracy	Precision	Recall	$F_{1}$ -Score	Precision	Recall	$F_{1}$ -Score
SuperPoint+LightGlue	87.36%	62.69%	68.82%	62.83%	93.67%	87.36%	89.93%
SIFT+FLANN	12.34%	7.29%	7.03%	6.39%	12.77%	12.34%	11.22%

TABLE I: Frame-to-panorama localization performance comparison between the proposed SuperPoint-LightGlue pipeline and the traditional SIFT-FLANN baseline. The task is evaluated as spatial-bin classification over the reference panorama.

Fig. 3: Examples of GMM clusters obtained from tile-level visual embeddings. The left cluster contains vessel-relevant tiles, while the right cluster contains visually similar but non-relevant structures such as waves, masts, and breakwater regions.

Fig. 4: Example of a manually curated segmentation annotation used to create object-detection labels. The annotation protocol focuses on visible vessel structures while excluding thin or ambiguous elements such as masts and railings.

Fold	Precision	Recall	F1-score	$\mathbf{AP_{50}}$	$\mathbf{AP_{50\textnormal{--}95}}$
Fold 1	0.9438	0.9015	0.9222	0.9485	0.7485
Fold 2	0.9599	0.8928	0.9252	0.9500	0.7623
Fold 3	0.9339	0.9092	0.9214	0.9527	0.7647
Fold 4	0.9532	0.8738	0.9117	0.9392	0.7221
Fold 5	0.9566	0.9067	0.9310	0.9486	0.7574
Mean $\boldsymbol{\pm}$ SD	$\mathbf{0.9495\pm 0.0106}$	$\mathbf{0.8968\pm 0.0143}$	$\mathbf{0.9223\pm 0.0070}$	$\mathbf{0.9478\pm 0.0051}$	$\mathbf{0.7510\pm 0.0173}$

TABLE II: Sequence-grouped five-fold outer cross-validation performance of the fine-tuned YOLO26-m detector. Each outer evaluation fold contains 44 images. The final row reports the unweighted mean

\pm

sample standard deviation across folds.

Fig. 6: Example prediction from the live PTZ camera feed using the trained YOLO26-m detector. All boats within the marina are detected as well as a small boat on the horizon outside the marina.

Explore similar work

Dec 10, 2025cs.CV

LiM-YOLO: Less is More with Pyramid Level Shift for Ship Detection in Optical Remote Sensing

General-purpose object detectors face fundamental structural limitations when applied to ship detection in satellite imagery, where the ship scale distribution is concentrated at small sizes and high aspect ratios. In conventional You Only Look Once architectures, the deepest feature pyramid level (stride 32) compresses narrow vessels into sub-pixel representations, causing severe spatial feature dilution and compromising accurate ship boundary regression. We propose Less is More YOLO, a streamlined detector built upon the extra-large variant of YOLOv9, to address these domain-specific structural conflicts. From a statistical analysis of ship scale distributions across four major benchmarks (SODA-A, DOTA-v1.5, FAIR1M-v2.0, and ShipRSImageNet), we introduce a Pyramid Level Shift Strategy that shifts the detection head from strides 8, 16, and 32 to strides 4, 8, and 16. This shift satisfies a spatial representability condition derived from the Nyquist-Shannon principle for the narrowest targets, while eliminating the computational redundancy of the deepest pyramid level. To further stabilize training on high-resolution satellite inputs, we incorporate a group-normalized composite-backbone projection module, mitigating gradient instability in memory-constrained micro-batch regimes. Validated on these four datasets, our detector attains an mAP50:95 of 0.600 with only 21.16 million parameters, a 64.1% reduction from the extra-large YOLOv9 baseline (58.99 million). Despite this compact size, our model surpasses state-of-the-art detectors up to three times larger, validating that a well-targeted pyramid level shift achieves a "Less is More" balance between accuracy and efficiency. The code is available at https://github.com/egshkim/LiM-YOLO.

Seon-Hoon Kim, Yerin Kim, Hyeji Sim +3

University of Science and Technology (UST), Daejeon 34113, Republic of Korea · Korea Aerospace Research Institute (KARI), Daejeon 34133, Republic of Korea

Ship Detection Yolo26 Satellite Imagery Remote Sensing

Ask PDF

Sep 23, 2026cs.CV

Strip Convolution and Direction-Aware Exclusion Loss for Oriented Ship Detection

Oriented ship detection in very high resolution (VHR) remote sensing imagery remains challenging due to elongated hull geometry and dense target distributions in complex port scenes. Existing methods typically address geometric representation and duplicate suppression separately. To jointly tackle these issues, we propose an oriented ship detector with two complementary components. The C3k2_Strip module employs orthogonal strip convolutions to better capture elongated hull structures, while the Class-Aware Direction-Aware Exclusion Loss (CA-DAEL) suppresses redundant predictions using class, direction, and confidence cues. Experiments on HRSC2016 and DIOR-R achieve 78.45% and 53.71% mAP50:95, respectively, with only 2.91M parameters. On HRSC2016, the proposed method improves mAP50:95 by 6.32 percentage points over the YOLOv11-OBB baseline, demonstrating its effectiveness for accurate oriented ship detection.

Bin Chen, Yuanyuan Liu, Peng Yang +1

School of Information and Software Engineering, East China Jiaotong University, Nanchang 330013, China. · Jiangxi Vocational University of Foreign Studies, Nanchang 330099, China.

Ship Detection Convolutional

Ask PDF

Jun 19, 2026cs.SD

Sea-Scan: High-Accuracy, ML-based Dark Vessel Detection and Localisation via Weakly Supervised DAS Monitoring

We present an ML-based vessel detection and localization system, trained with weak supervision from imperfect AIS labels, that achieves a 97.8% detection rate at 1.98% false-trigger rate, successfully identifies dark-vessel events from unlabeled data.

Tian Tian, Agastya Raj, Lara Flanagan +2

School of Computer Science and Statistics, IRIS Research Group, ADAPT Research Centre, Trinity College Dublin, Ireland · School of Engineering, ADAPT Research Centre, Trinity College Dublin, Ireland

Ship Detection Automatic Identification System Localization

Ask PDF