跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2603.25351 · v2

Image Rotation Angle Estimation: Five Circular-Aware Methods Across Sixteen Architectures

首次提交 2026-03-26 19:57
版本更新 2026-11-15 05:27
雷达观测 2026-11-17 12:06
arXiv ↗PDF ↗

摘要

Automatic image rotation estimation is a key preprocessing step in many vision pipelines, yet challenging because angles have circular topology, creating boundary discontinuities that hinder standard regression. We present a comprehensive study of five circular-aware methods for global orientation estimation: three regression based (direct angle with circular loss, unit vector, phase-shifting coder) and two classification based (angular binning, circular Gaussian distribution). Using transfer learning from ImageNet-pretrained models, we evaluate them across sixteen modern architectures and two data regimes of natural images, small-scale DRC-D and large-scale COCO. Classification approaches are the stronger family, attaining the lowest error on thirteen of the sixteen DRC-D backbones. The decisive ingredient is supervision that encodes angular neighborhood structure: plain cross-entropy lacks it and fails intermittently, while circular smooth labels (CSL) eliminate every observed training failure and reach the lowest DRC-D error, 0.81° mean absolute error (MAE). On COCO, the common rotate-and-crop protocol leaks the applied angle through the rotated JPEG block grid, a shortcut that the classification heads learned and the regression heads largely did not. Re-encoding every training view after rotation suppresses it. Under this trace-free protocol and on re-encoded test views, classification again leads on all three backbones tested, and the circular Gaussian distribution (CGD) with ConvNeXt V2 Base reaches 3.76° on COCO 2017 (three-run mean), against 6.09° for the strongest prior model on the same views. CSL and CGD are also the cheapest methods to train.

作者备注:20 pages, 5 figures, 9 tables. v2: The COCO experiments were repeated under a trace-free protocol, after we found that the rotate-and-crop protocol leaks the angle through the JPEG block grid. Title changed. Code, model and demo: https://github.com/maxwoe/image-rotation-angle-estimation

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

本站首次收录为 v2,暂无更早版本可对照。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v2 · 2026-11-15 05:27 · 最新观测版本

Image Rotation Angle Estimation: Five Circular-Aware Methods Across Sixteen Architectures

Automatic image rotation estimation is a key preprocessing step in many vision pipelines, yet challenging because angles have circular topology, creating boundary discontinuities that hinder standard regression. We present a comprehensive study of five circular-aware methods for global orientation estimation: three regression based (direct angle with circular loss, unit vector, phase-shifting coder) and two classification based (angular binning, circular Gaussian distribution). Using transfer learning from ImageNet-pretrained models, we evaluate them across sixteen modern architectures and two data regimes of natural images, small-scale DRC-D and large-scale COCO. Classification approaches are the stronger family, attaining the lowest error on thirteen of the sixteen DRC-D backbones. The decisive ingredient is supervision that encodes angular neighborhood structure: plain cross-entropy lacks it and fails intermittently, while circular smooth labels (CSL) eliminate every observed training failure and reach the lowest DRC-D error, 0.81° mean absolute error (MAE). On COCO, the common rotate-and-crop protocol leaks the applied angle through the rotated JPEG block grid, a shortcut that the classification heads learned and the regression heads largely did not. Re-encoding every training view after rotation suppresses it. Under this trace-free protocol and on re-encoded test views, classification again leads on all three backbones tested, and the circular Gaussian distribution (CGD) with ConvNeXt V2 Base reaches 3.76° on COCO 2017 (three-run mean), against 6.09° for the strongest prior model on the same views. CSL and CGD are also the cheapest methods to train.

查看此版本解读 →查看此版本原文 ↗