Cascade RetinaNet: Maintaining Consistency for Single-Stage Object Detection

TL;DR

This paper introduces Cas-RetinaNet, a multistage object detector that enhances single-stage detection by improving consistency in classification and localization, leading to significant performance gains on MS COCO.

Contribution

The paper proposes a novel cascade framework with a Feature Consistency Module to address misalignment issues in single-stage detectors, achieving improved accuracy.

Findings

Cas-RetinaNet improves AP from 39.1 to 41.1 on MS COCO.

The method maintains stable performance across different models and scales.

Addressing inconsistency is key to enhancing single-stage detection performance.

Abstract

Recent researches attempt to improve the detection performance by adopting the idea of cascade for single-stage detectors. In this paper, we analyze and discover that inconsistency is the major factor limiting the performance. The refined anchors are associated with the feature extracted from the previous location and the classifier is confused by misaligned classification and localization. Further, we point out two main designing rules for the cascade manner: improving consistency between classification confidence and localization performance, and maintaining feature consistency between different stages. A multistage object detector named Cas-RetinaNet, is then proposed for reducing the misalignments. It consists of sequential stages trained with increasing IoU thresholds for improving the correlation, and a novel Feature Consistency Module for mitigating the feature inconsistency. Experiments show that our proposed Cas-RetinaNet achieves stable performance gains across different models and input scales. Specifically, our method improves RetinaNet from 39.1 AP to 41.1 AP on the challenging MS COCO dataset without any bells or whistles.