BoCS: Image retrieval using explicable methods

Author(s)

Silva, Endi D. C. ; Traina, Agma J. M.

• Document type: Journal article
• Source: 2023 36TH CONFERENCE ON GRAPHICS, PATTERNS AND IMAGES, SIBGRAPI 2023; v. N/A, p. 6-pg., 2023-01-01.

Abstract

In recent years, image generation has been growing at a very fast pace, demanding specific systems for managing large image datasets. For example, we can mention the content-based image retrieval (CBIR) systems. Usually, they use a representation of feature vectors based on the images' visual content to store/retrieve them and to perform demanded queries. Currently, neural networks perform the task of generating image representations with great mastery. However, these networks usually create methods that are difficult to understand or to explain, which for some applications, such as medical decision-making systems, can be a significant disadvantage. Thinking about the explainability aspect, in this work, we present a new technique based on the bag of visual words (BoVW) which, in addition to generating promising explainable methods, has long been the state of the art for generating image representations. The results showed that the presented method BoCS overcomes similar methods and still has the potential to be further explored. (AU)

• FAPESP's process: 16/17078-0 - Mining, indexing and visualizing Big Data in clinical decision support systems (MIVisBD)
• Grantee: Agma Juci Machado Traina
• Support Opportunities: Research Projects - Thematic Grants