Matches in Nanopublications for { <https://hal-lirmm.ccsd.cnrs.fr/lirmm-03815628/document> ?p ?o ?g. }
Showing items 1 to 12 of
12
with 100 items per page.
- document type Paper assertion.
- document type Resource assertion.
- document type MediaObject assertion.
- document description "The multiplication of publicly available datasets makes it possible to develop Deep Learning models for many real-world applica tions. However, some domains are still poorly explored, and their related datasets are often small or inconsistent. In addition, some biases linked to the dataset construction or labeling may give the impression that a model is particularly efficient. Therefore, evaluating a model requires a clear understanding of the database. Moreover, a model often reflects a given dataset’s performance and may deteriorate if a shift exists between the training dataset and real-world data. In this paper, we derive a more consistent and balanced version of the TrashCan [6] image dataset, called UNO, to evaluate models for de tecting non-natural objects in the underwater environment. We pro pose a method to balance the number of annotations and images for cross-evaluation. We then compare the performance of a SOTA object detection model when using TrashCAN and UNO datasets. Addition ally, we assess covariate shift by testing the model on an image dataset for real-world application. Experimental results show significantly better and more consistent performance using the UNO dataset. The UNO database and the code are publicly av https://github.com/CBarrelet/balanced_kfold" assertion.
- document dateCreated "2025-06-11 11:05:31.499118+00:00" assertion.
- document name "From TrashCan to UNO: Deriving an Underwater Image Dataset To Get a More Consistent and Balanced Version" assertion.
- document contentUrl "https://hal-lirmm.ccsd.cnrs.fr/lirmm-03815628/document" assertion.
- document creator mailto:taha.lahami@ve.ismar.cnr.it assertion.
- document dateModified "2025-10-14 08:34:03.226349+00:00" assertion.
- document license legalcode assertion.
- document sdDatePublished "2025-06-11 11:05:31.499118+00:00" assertion.
- document author mailto:taha.lahami@ve.ismar.cnr.it assertion.