An Annotated and Classified Maritime Dataset aimed at Machine Learning

In the context of the MSc dissertation "The Multi-Object Tracking with Multimodal Information for Autonomous Surface Vehicles", an annotated and classified maritime dataset was created in order to train a machine learning object detector. This dataset combines images from open-source datasets (such as the Singapore Maritime Dataset and the Kaggle Boat Types Recognition Dataset) while adding data collected with the Sense ASV in two portuguese ports (Leixões and Viana do Castelo). The content is reorganised in 9 categories (ore carrier, bulk carrier, container ship, cruise ship, ferry boat, sail boat, fishing boat, small boat, uncategorized), grouping up some less represented categories from the previous datasets with similar characteristics. The resulting dataset, contains 9044 images and their respective 21170 annotations.

Data and Resources

Additional Info

Field Value
Source Singapore Maritime Dataset; Boat types recognition (Kaggle, v.1); SeaShips: A Large-Scale Precisely Annotated Dataset for Ship Detection (DOI: 10.1109/TMM.2018.2865686)
Author Diogo Duarte
Last Updated February 17, 2023, 08:55 (UTC)
Created January 20, 2022, 11:45 (UTC)
Citation Duarte, D., Pereira, M. I., & Pinto, A. M. (2022). An Annotated and Classified Maritime Dataset aimed at Machine Learning [Data set]. INESC TEC.
dc.Contributor Maria Inês Pereira; Andry Maykol Pinto
dc.Coverage.Spatial Singapore port; Leixões and Viana do Castelo ports.
dc.Coverage.Temporal October 2017 - November 2021
dc.Created.Date January 17th, 2022
dc.Format .zip; *.jpg; *txt
dc.Language EN
dc.Relation Multiple Vessel Detection and Tracking in Harsh Maritime Environments. Oceans Maritime Conference 2021 (To appear).
dc.Type Images form on-board camera (.jpg), annotated objects in YOLO format (.txt)
ddi.InstrumentName Mynteye
ddi.Software ROS (python) (C++) - to collect data from the SenseASV
ddi.TypeInstrument Video camera