RECONSURVE's Boats Dataset


The RECONSURVE project has, as a key objective, to recognize different category of boats in European waters. One of the difficulties when we try to address this problem is the absence of a well designed and annotated image dataset. Such datasets are mandatory for training recognition algorithms and evaluating their performance. For this reason, we have created the RECONSURVE image databas. The database contains 17 boat categories, and a total of 5303 images. The number of images varies across categories, but there are at least 100 images per category. Within the 17 categories, we consider two different types of settings/experiments, corresponding to two different use cases: (a) on one hand, 5 main categories (Boating, Cargo, Passenger, Fishing and Tanker), and (b) on the other hand the remain categories are considered as sub-categories (e.g. petrolier is also considered as tanker, ligneur is also considered as fishing).

Attributes: the dataset also includes annotations describing boat attributes (cranes, fishing rodes, etc.). Bounding boxes are given.

Download information

The dataset can be downloaded here: [Download link]

Database citation

Ronan Sicre, Frédéric Jurie. Discriminative part model for visual recognition. Technical Report, GREYC CNRS UMR 6072, Universite de Caen, hal-01132389, 2015 [pdf]


Questions, comments etc. about the database are welcome.