🤖 AI Summary
Conventional superpixel segmentation methods assume planar geometry and are ill-suited for 360° spherical images, lacking dedicated approaches that respect spherical manifold structure. Method: This paper proposes the first end-to-end deep learning framework for spherical superpixel segmentation. It employs spherical convolutional neural networks to model the intrinsic geometry of the sphere, integrates differentiable K-means to enforce spherical constraints during clustering, and introduces spherical-coordinate-aware data augmentation alongside manifold-aware feature learning. Contribution/Results: The method explicitly incorporates spherical geometry—departing fundamentally from planar assumptions—and constitutes the first deep superpixel segmentation approach tailored to spherical imagery. Evaluated on two spherical image datasets, it significantly outperforms state-of-the-art planar and spherical superpixel methods, demonstrating that explicit spherical geometric modeling critically enhances segmentation accuracy.
📝 Abstract
Over the years, the use of superpixel segmentation has become very popular in various applications, serving as a preprocessing step to reduce data size by adapting to the content of the image, regardless of its semantic content. While the superpixel segmentation of standard planar images, captured with a 90{deg} field of view, has been extensively studied, there has been limited focus on dedicated methods to omnidirectional or spherical images, captured with a 360{deg} field of view. In this study, we introduce the first deep learning-based superpixel segmentation approach tailored for omnidirectional images called DSS (for Deep Spherical Superpixels). Our methodology leverages on spherical CNN architectures and the differentiable K-means clustering paradigm for superpixels, to generate superpixels that follow the spherical geometry. Additionally, we propose to use data augmentation techniques specifically designed for 360{deg} images, enabling our model to efficiently learn from a limited set of annotated omnidirectional data. Our extensive validation across two datasets demonstrates that taking into account the inherent circular geometry of such images into our framework improves the segmentation performance over traditional and deep learning-based superpixel methods. Our code is available online.