Learning Modulated Loss for Rotated Object Detection

Abstract

Popular rotated detection methods usually use five parameters (coordinates of the central point, width, height, and rotation angle) to describe the rotated bounding box and l1-loss as the loss function. In this paper, we argue that the aforementioned integration can cause training instability and performance degeneration, due to the loss discontinuity resulted from the inherent periodicity of angles and the associated sudden exchange of width and height. This problem is further pronounced given the regression inconsistency among five parameters with different measurement units. We refer to the above issues as rotation sensitivity error (RSE) and propose a modulated rotation loss to dismiss the loss discontinuity. Our new loss is combined with the eight-parameter regression to further solve the problem of inconsistent parameter regression. Experiments show the state-of-art performances of our method on the public aerial image benchmark DOTA and UCAS-AOD. Its generalization abilities are also verified on ICDAR2015, HRSC2016, and FDDB. Qualitative improvements can be seen in Fig 1, and the source code will be released with the publication of the paper.

11 pages, 8 figures

References in corpus (13)

• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
• R-FCN: Object Detection via Region-based Fully Convolutional Networks
• Object Detection in Optical Remote Sensing Images: A Survey and A New Benchmark
• DSSD : Deconvolutional Single Shot Detector
• OverFeat: Integrated Recognition, Localization and Detection using Convolutional Networks
• TextBoxes++: A Single-Shot Oriented Scene Text Detector
• Automatic Ship Detection of Remote Sensing Images from Google Earth in Complex Scenes Based on Multi-Scale Rotation Dense Feature Pyramid Networks
• R2CNN: Rotational Region CNN for Orientation Robust Scene Text Detection
• TextBoxes: A Fast Text Detector with a Single Deep Neural Network
• Hybrid Task Cascade for Instance Segmentation
• Learning a Rotation Invariant Detector with Rotatable Bounding Box
• R3Det: Refined Single-Stage Detector with Feature Refinement for Rotating Object
• Omnidirectional Scene Text Detection with Sequential-free Box Discretization

Cited by in corpus (14)

• Anchor-free Oriented Proposal Generator for Object Detection
• Optimization for Arbitrary-Oriented Object Detection via Representation Invariance Loss
• PolarDet: A Fast, More Precise Detector for Rotated Target in Aerial Images
• Arbitrary-Oriented Ship Detection through Center-Head Point Extraction
• LO-Det: Lightweight Oriented Object Detection in Remote Sensing Images
• DARDet: A Dense Anchor-free Rotated Object Detector in Aerial Images
• Dense Label Encoding for Boundary Discontinuity Free Rotation Detection
• Rotated Object Detection via Scale-invariant Mahalanobis Distance in Aerial Images
• MRDet: A Multi-Head Network for Accurate Oriented Object Detection in Aerial Images
• RAPiD: Rotation-Aware People Detection in Overhead Fisheye Images
• Sampling Equivariant Self-attention Networks for Object Detection in Aerial Images
• OSKDet: Towards Orientation-sensitive Keypoint Localization for Rotated Object Detection
• Towards Toxic and Narcotic Medication Detection with Rotated Object Detector
• MidNet: An Anchor-and-Angle-Free Detector for Oriented Ship Detection in Aerial Images