Self-Balanced R-CNN for Instance Segmentation

TL;DR

This paper introduces Self-Balanced R-CNN, a novel instance segmentation model that addresses training imbalances and feature-level issues, achieving state-of-the-art results with fewer parameters and improved accuracy.

Contribution

It proposes new loop mechanisms, an improved RoI extraction, and a fully convolutional architecture head to enhance instance segmentation performance.

Findings

Achieves 45.3% AP for detection and 41.5% AP for segmentation on COCO.

Addresses IoU distribution imbalance and feature-level imbalance effectively.

Reduces model complexity while maintaining or improving accuracy.

Abstract

Current state-of-the-art two-stage models on instance segmentation task suffer from several types of imbalances. In this paper, we address the Intersection over the Union (IoU) distribution imbalance of positive input Regions of Interest (RoIs) during the training of the second stage. Our Self-Balanced R-CNN (SBR-CNN), an evolved version of the Hybrid Task Cascade (HTC) model, brings brand new loop mechanisms of bounding box and mask refinements. With an improved Generic RoI Extraction (GRoIE), we also address the feature-level imbalance at the Feature Pyramid Network (FPN) level, originated by a non-uniform integration between low- and high-level features from the backbone layers. In addition, the redesign of the architecture heads toward a fully convolutional approach with FCC further reduces the number of parameters and obtains more clues to the connection between the task to solve and the layers used. Moreover, our SBR-CNN model shows the same or even better improvements if adopted in conjunction with other state-of-the-art models. In fact, with a lightweight ResNet-50 as backbone, evaluated on COCO minival 2017 dataset, our model reaches 45.3% and 41.5% AP for object detection and instance segmentation, with 12 epochs and without extra tricks. The code is available at https://github.com/IMPLabUniPr/mmdetection/tree/sbr_cnn