CVSep 6, 2018

Labeling Panoramas with Spherical Hourglass Networks

arXiv:1809.02123v13 citations
Originality Synthesis-oriented
AI Analysis

This addresses the problem of visual perception in spherical cameras for applications like consumer 360° imaging, though it appears incremental as it adapts existing hourglass networks to a spherical domain.

The paper tackles dense labeling of 360° panoramic images by introducing a spherical convolutional hourglass network (SCHN), which is invariant to camera orientation and scalable, showing promising results on a spherical semantic segmentation task.

With the recent proliferation of consumer-grade 360° cameras, it is worth revisiting visual perception challenges with spherical cameras given the potential benefit of their global field of view. To this end we introduce a spherical convolutional hourglass network (SCHN) for the dense labeling on the sphere. The SCHN is invariant to camera orientation (lifting the usual requirement for `upright' panoramic images), and its design is scalable for larger practical datasets. Initial experiments show promising results on a spherical semantic segmentation task.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes