MARITIME IMAGES ANNOTATION

Annotate and classify  OSCAR‘s training data (millions of images), a product that uses artificial intelligence to visually identify and classify objects in water during navigation in busy maritime areas.

CHALLENGE

Annotate the training data of a product that uses artificial intelligence to visually identify and classify objects in water during navigation in a busy maritime area.

For a complete training set, it was estimated that it’s necessary to annotate more than 1 000 000 images, between thermal and color images, with precision and accuracy, following a classification tree with more than 50 branches.

SOLUTION

To meet a challenge of this magnitude, a strategy was defined to meet the deadlines and requirements of the client.
This plan lasts one year and six months, with one plateau in the 800-hour annotation per month so that, on average, 60,000 annotated images per month can be achieved.

Periodically, a neuronal network training is done, by the client, with the available dataset to evaluate the results and progress. On our part, a weekly follow-up is provided with detailed information, such as the number of objects annotated, classifications, among other parameters.

Found Something interesting?

Get in touch with our team, we will be glad to assist you.

Diogo Arreda

Business Developer