Título: Novelty detection in underwater inspection videos
Autores: Evelyn C. S. Batista, Wouter Caarls, Leonardo A. Mendoza, Marco Aurelio Pacheco
Resumo: This work consists of a study on the detection of anomalies in underwater inspection videos. Novelty detection is the identification of new or distinct data from a dataset, where the challenge is for an intelligent algorithm to be able to detect an input pattern as being previously unknown. Then, different models with different configurations were chosen, trained using different video frames and analyzed in order to find the best pattern detection model. The objective of this work is to contribute to the automatic detection of anomalies, which is currently a task carried out by specialists who analyze these videos. The model shows promise compared to other similar works in the area.
Palavras-chave: Novelty Detection, Anomaly Detection, Autoencoder, Deep Learning
Páginas: 7
Código DOI: 10.21528/CBIC2023-037
Artigo em pdf: CBIC_2023_paper037.pdf
Arquivo BibTeX: CBIC_2023_037.bib