2019/10/02 by Daniel Ruíz, Ruiz, Daniel V., Bruno A. Krinski +3 · 1 citation
Computer Science · #Advanced Image and Video Retrieval Techniques #Advanced Neural Network Applications #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Visual Attention and Saliency Detection
paper · pdf · doi:10.48550/arxiv.1910.01256
openalex publication_date 2019/10/02 · openalex created_date 2022/07/28 · openalex updated_date 2026/07/28
In this paper, we propose a novel data augmentation technique (ANDA) applied\nto the Salient Object Detection (SOD) context. Standard data augmentation\ntechniques proposed in the literature, such as image cropping, rotation,\nflipping, and resizing, only generate variations of the existing examples,\nproviding a limited generalization. Our method has the novelty of creating new\nimages, by combining an object with a new background while retaining part of\nits salience in this new context; To do so, the ANDA technique relies on the\nlinear combination between labeled salient objects and new backgrounds,\ngenerated by removing the original salient object in a process known as image\ninpainting. Our proposed technique allows for more precise control of the\nobject's position and size while preserving background information. Aiming to\nevaluate our proposed method, we trained multiple deep neural networks and\ncompared the effect that our technique has in each one. We also compared our\nmethod with other data augmentation techniques. Our findings show that\ndepending on the network improvement can be up to 14.1% in the F-measure and\ndecay of up to 2.6% in the Mean Absolute Error.\n