Hyperspectral Image Dataset for Benchmarking on Salient Object Detection

TL;DR

This paper introduces a new hyperspectral image dataset specifically designed for benchmarking salient object detection, addressing the lack of dedicated datasets and evaluating existing models on this new benchmark.

Contribution

The paper provides a publicly available hyperspectral salient object detection dataset with 60 images, ground-truth labels, and color renderings, enabling standardized evaluation.

Findings

Existing models achieve varying AUC scores on the new dataset.

The dataset includes diverse object sizes and contrast conditions.

Benchmark results highlight the need for improved hyperspectral saliency models.

Abstract

Many works have been done on salient object detection using supervised or unsupervised approaches on colour images. Recently, a few studies demonstrated that efficient salient object detection can also be implemented by using spectral features in visible spectrum of hyperspectral images from natural scenes. However, these models on hyperspectral salient object detection were tested with a very few number of data selected from various online public dataset, which are not specifically created for object detection purposes. Therefore, here, we aim to contribute to the field by releasing a hyperspectral salient object detection dataset with a collection of 60 hyperspectral images with their respective ground-truth binary images and representative rendered colour images (sRGB). We took several aspects in consideration during the data collection such as variation in object size, number of objects, foreground-background contrast, object position on the image, and etc. Then, we prepared ground truth binary images for each hyperspectral data, where salient objects are labelled on the images. Finally, we did performance evaluation using Area Under Curve (AUC) metric on some existing hyperspectral saliency detection models in literature.