Non-local RoIs for Instance Segmentation

TL;DR

This paper introduces Non-Local RoI (NL-RoI) Blocks that enhance Mask R-CNN by allowing object regions to incorporate information from all other regions, improving instance segmentation performance.

Contribution

The paper proposes a novel NL-RoI Block module that models inter-object relationships within Mask R-CNN, providing a flexible and effective enhancement.

Findings

NL-RoI Blocks improve Mask R-CNN performance on Robust Vision Challenge benchmarks.

The module is low-cost and easily adaptable to different Mask R-CNN heads.

Experimental results demonstrate significant gains in instance segmentation accuracy.

Abstract

We introduce the concept of Non-Local RoI (NL-RoI) Block as a generic and flexible module that can be seamlessly adapted into different Mask R-CNN heads for various tasks. Mask R-CNN treats RoIs (Regions of Interest) independently and performs the prediction based on individual object bounding boxes. However, the correlation between objects may provide useful information for detection and segmentation. The proposed NL-RoI Block enables each RoI to refer to all other RoIs' information, and results in a simple, low-cost but effective module. Our experimental results show that generalizations with NL-RoI Blocks can improve the performance of Mask R-CNN for instance segmentation on the Robust Vision Challenge benchmarks.