Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Image Rotation Angle Estimation: Comparing Circular-Aware Methods

Maximilian Woehrer

Abstract

Automatic image rotation estimation is a key preprocessing step in many vision pipelines. This task is challenging because angles have circular topology, creating boundary discontinuities that hinder standard regression methods. We present a comprehensive study of five circular-aware methods for global orientation estimation: direct angle regression with circular loss, classification via angular binning, unit-vector regression, phase-shifting coder, and circular Gaussian distribution. Using transfer learning from ImageNet-pretrained models, we systematically evaluate these methods across sixteen modern architectures by adapting their output heads for rotation-specific predictions. Our results show that probabilistic methods, particularly the circular Gaussian distribution, are the most robust across architectures, while classification achieves the best accuracy on well-matched backbones but suffers training instabilities on others. The best configuration (classification with EfficientViT-B3) achieves a mean absolute error (MAE) of 1.23° (mean across five independent runs) on the DRC-D dataset, while the circular Gaussian distribution with MambaOut Base achieves a virtually identical 1.24° with greater robustness across backbones. Training and evaluating our top-performing method-architecture combinations on COCO 2014, the best configuration reaches 3.71° MAE, improving substantially over prior work, with further improvement to 2.84° on the larger COCO 2017 dataset.

Image Rotation Angle Estimation: Comparing Circular-Aware Methods

Abstract

Automatic image rotation estimation is a key preprocessing step in many vision pipelines. This task is challenging because angles have circular topology, creating boundary discontinuities that hinder standard regression methods. We present a comprehensive study of five circular-aware methods for global orientation estimation: direct angle regression with circular loss, classification via angular binning, unit-vector regression, phase-shifting coder, and circular Gaussian distribution. Using transfer learning from ImageNet-pretrained models, we systematically evaluate these methods across sixteen modern architectures by adapting their output heads for rotation-specific predictions. Our results show that probabilistic methods, particularly the circular Gaussian distribution, are the most robust across architectures, while classification achieves the best accuracy on well-matched backbones but suffers training instabilities on others. The best configuration (classification with EfficientViT-B3) achieves a mean absolute error (MAE) of 1.23° (mean across five independent runs) on the DRC-D dataset, while the circular Gaussian distribution with MambaOut Base achieves a virtually identical 1.24° with greater robustness across backbones. Training and evaluating our top-performing method-architecture combinations on COCO 2014, the best configuration reaches 3.71° MAE, improving substantially over prior work, with further improvement to 2.84° on the larger COCO 2017 dataset.

Paper Structure

This paper contains 22 sections, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Circular-Aware Methods
  4. Regression Methods
  5. Direct Angle with Circular Loss
  6. Unit Vector Approach
  7. Phase-Shifting Coder
  8. Classification Methods
  9. Classification
  10. Circular Gaussian Distribution
  11. Experiments
  12. Datasets
  13. Rotation Methodology
  14. Transfer Learning Framework
  15. Training and Optimization
  16. ...and 7 more sections

Figures (2)

  • Figure 1: Model architecture. A pretrained backbone extracts features from the rotated input, which are passed to one of five circular-aware angle heads to predict the rotation angle $\hat{\theta}$.
  • Figure 2: Qualitative results of CGD (MambaOut Base) on COCO 2014. Left six: accurate predictions; right six: failure cases.