MODS—A USV-Oriented Object Detection and Obstacle Segmentation Benchmark

Bovcon, Borja; Muhovič, Jon; Vranac, Duško; Mozetič, Dean; Perš, Janez; Kristan, Matej

doi:10.1109/tits.2021.3124192

Cited by 53 publications

(53 citation statements)

References 56 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The proposed scaffolding learning regime (SLR) is evaluated on the most recent maritime obstacle detection benchmark MODS [9], which contains approximately 100 annotated sequences captured under various conditions. The evaluation protocol reflects the detection performance meaningful for practical USV navigation and separately evaluates (i) the accuracy of obstacle-water edge estimation for static obstacles and (ii) the detection performance for dynamic obstacles.…”

Section: Resultsmentioning

confidence: 99%

“…Features are in turn used to refine the unlabeled regions of the constraints-generated partial labels. Experimental results on the currently most challenging maritime obstacle detection dataset [9] show that models trained using SLR outperform models classically trained from full dense annotations, which is a remarkable result. To the best of our knowledge, this is the first method for training obstacle detection from weak annotations in the marine domain which surpasses fully supervised training from dense labels.…”

Section: Introductionmentioning

confidence: 92%

“…Finally, we compare the performance of the SLR-trained networks and classically-trained state-of-the-art methods from [9] (RefineNet [25], DeepLabV3 [11], BiSeNet [42] and WaSR [6]). In addition to WaSR, we also demonstrate SLR on DeepLabV3 -in the following, we refer to these two networks as WaSR SLR and DeepLabV3 SLR .…”

Section: Comparison With the State-of-the-artmentioning

confidence: 99%

“…But falsely classifying isolated patches of water pixels in front of the boat as obstacles will detrimentally affect the control, causing frequent unnecessary stops. This is reflected in recent marine benchmarks [9], where performance is evaluated in terms of dynamic obstacle detection and water-edge estimation accuracy, while the segmentation accuracy beyond the water-obstacle boundary is ignored.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Learning Maritime Obstacle Detection from Weak Annotations by Scaffolding

Žust¹,

Kristan²

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Coastal water autonomous boats rely on robust perception methods for obstacle detection and timely collision avoidance. The current state-of-the-art is based on deep segmentation networks trained on large datasets. Per-pixel ground truth labeling of such datasets, however, is laborintensive and expensive. We observe that far less information is required for practical obstacle avoidance -the location of water edge on static obstacles like shore and approximate location and bounds of dynamic obstacles in the water is sufficient to plan a reaction. We propose a new scaffolding learning regime (SLR) that allows training obstacle detection segmentation networks only from such weak annotations, thus significantly reducing the cost of ground-truth labeling. Experiments show that maritime obstacle segmentation networks trained using SLR substantially outperform the same networks trained with dense ground truth labels. Thus accuracy is not sacrificed for labelling simplicity but is in fact improved, which is a remarkable result.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 92%

Section: Comparison With the State-of-the-artmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Learning Maritime Obstacle Detection from Weak Annotations by Scaffolding

Žust¹,

Kristan²

2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Cameras as low-power and information rich sensors are particularly appealing due to their large success in perception for autonomous cars [1], [2]. However, recent works [3], [4] have shown that methods developed for autonomous cars do not translate well to USVs due to the specifics of the maritime domain. As a result, several approaches that exploit the domain specifics for improved detection accuracy have been recently proposed [5], [6], [7], [8].…”

Section: Introductionmentioning

confidence: 99%

Temporal Context for Robust Maritime Obstacle Detection

Žust¹,

Kristan²

2022

Preprint

Self Cite

View full text Add to dashboard Cite

Robust maritime obstacle detection is essential for fully autonomous unmanned surface vehicles (USVs). The currently widely adopted segmentation-based obstacle detection methods are prone to misclassification of object reflections and sun glitter as obstacles, producing many false positive detections, effectively rendering the methods impractical for USV navigation. However, water-turbulence-induced temporal appearance changes on object reflections are very distinctive from the appearance dynamics of true objects. We harness this property to design WaSR-T, a novel maritime obstacle detection network, that extracts the temporal context from a sequence of recent frames to reduce ambiguity. By learning the local temporal characteristics of object reflection on the water surface, WaSR-T substantially improves obstacle detection accuracy in the presence of reflections and glitter. Compared with existing single-frame methods, WaSR-T reduces the number of false positive detections by 41% overall and by over 53% within the danger zone of the boat, while preserving a high recall, and achieving new state-of-the-art performance on the challenging MODS maritime obstacle detection benchmark.

show abstract

Field experiment of autonomous ship navigation in canal and surrounding nearshore environments

Kim,

Lee,

Chung

et al. 2023

Journal of Field Robotics

View full text Add to dashboard Cite

In this paper, we present the development of autonomous navigation capabilities for small cruise boats, and their verification by field experiments in a canal and its surrounding waters. A cruise boat was converted to an autonomous surface vehicle (ASV) by installing various sensors and actuators to enable autonomous navigation. Navigation and perception sensors, such as global positioning system, attitude and heading reference system, radar, light detection and ranging (LiDAR), and cameras, were mounted on the ASV to estimate its motion and perceive the surrounding environment. Motors and potentiometers were installed for active control of the ASV. Software system components including navigation filters, object‐detection, path‐planning, and control algorithms were designed and implemented. In the narrow canal region, LiDARs were used to detect the side walls and boundaries of the canal. In open areas outside the canal, obstacles and object features were detected using various combinations of onboard sensors. A model‐based path‐planning algorithm was designed to avoid the detected obstacles, and the line‐of‐sight guidance was employed to control the vehicle. The performance of the developed system was verified through a field experiment in a real‐world maritime environment.

show abstract

MODS—A USV-Oriented Object Detection and Obstacle Segmentation Benchmark

Cited by 53 publications

References 56 publications

Learning Maritime Obstacle Detection from Weak Annotations by Scaffolding

Learning Maritime Obstacle Detection from Weak Annotations by Scaffolding

Temporal Context for Robust Maritime Obstacle Detection

Field experiment of autonomous ship navigation in canal and surrounding nearshore environments

Contact Info

Product

Resources

About