Computing a Characteristic Orientation for Rotation-Independent Image Analysis
Authors: Cristian Valero-Abundio, Emilio Sansano-Sansano, Raúl Montoliu, Marina Martínez García
Organizations: Institute of New Imaging Technologies, Universitat Jaume I, 12071 Castellón, Spain · Mathematics Department, Universitat Jaume I, 12071 Castellón, Spain
Handling geometric transformations, particularly rotations, remains a challenge in deep learning for computer vision. Standard neural networks lack inherent rotation invariance and typically rely on data augmentation or architectural modifications to improve robustness. Although effective, these approaches increase computational demands, require specialised implementations, or alter network structures, limiting their applicability. This paper introduces General Intensity Direction (GID), a preprocessing method that improves rotation robustness without modifying the network architecture. The method estimates a global orientation for each image and aligns it to a canonical reference frame, allowing standard models to process inputs more consistently across different rotations. Unlike moment-based approaches that extract invariant descriptors, this method directly transforms the image while preserving spatial structure, making it compatible with convolutional networks. Experimental evaluation on the rotated MNIST dataset shows that the proposed method achieves higher accuracy than state-of-the-art rotation-invariant architectures. Additional experiments on the CIFAR-10 dataset, confirm that the method remains effective under more complex conditions.
Figures & tables
Figure 1: Conv32 architecture diagram.
Figure 2: Upper row: mean accuracy (with 10 repetitions) across rotation angles from 0∘ to 359∘ for different model configurations. (a) results on rotated MNIST; (b) results on rotated CIFAR-10. Lower row: (c) mean accuracy (with 10 repetitions) for GID+Conv32 using different interpolation methods, evaluated across all rotation angles in the rotated MNIST dataset. The shaded area represents one standard deviation.
Methods
Input size
MNIST
RotMNIST
CNN
32 x 32
99.55%
45.42%
ORN [ 28 ]
32 x 32
99.42%
80.01%
RotEqNet [ 20 ]
28 x 28
99.26%
73.20%
G-CNN [ 1 ]
28 x 28
99.27%
44.81%
H-Net [ 26 ]
32 x 32
99.19%
92.44%
GA-CNN [ 9 ]
28 x 28
95.67%
93.29%
Table 1: The classification accuracies from various methods on the original test set and rotated test sets of MNIST. Bold and underline stand for best and 2nd best results. Results for the comparison are obtained from [ 21 ] .
Figure 3: Effect of the GID preprocessing method on MNIST (a) and CIFAR-10 (b). In each case, the first row shows original images, the second row shows the same inputs after a random rotation, and the third row shows the result after applying GID. While GID successfully recovers the same orientation in MNIST, its performance is less reliable in CIFAR-10 due to background interference.
Neural network robustness studies how models withstand adversarial inputs, data shifts, noise, and hardware faults while maintaining reliable predictions.
Multi-oriented text is ubiquitous in real-world scenes and remains a major challenge for scene text recognition (STR). Existing rotation-aware methods explicitly estimate text orientation. However, due to the lack of theoretical guarantees, they are prone to error accumulation, increased computational cost, and strong reliance on data. In this work, we incorporate rotation invariance into the STR framework to address these limitations. Specifically, we adopt an encoder-decoder architecture, embedding rotation equivariance in the encoder and rotation invariance in the decoder to construct a fully rotation-invariant network. On the decoder side, we first identify and prove the rotation-invariant property of the cross-attention mechanism and use it to formulate a rotation-invariant text decoder that maps visual features to output text in a rotation-invariant manner. On the encoder side, we propose a rotation-equivariant local-global extraction network that integrates deep equivariant convolutions with self-attention, enabling rotation-equivariant feature extraction while modeling inter-character dependencies and preserving fine-grained visual details. By integrating the encoder and decoder, we obtain an end-to-end Rotation-Invariant Scene Text Recognition network (RISTER). RISTER provides rotation invariance with theoretical guarantees, enhancing robustness on multi-oriented samples without introducing additional inference computation or relying on data-driven orientation correction. Experiments show that RISTER achieves state-of-the-art performance on both standard and multi-oriented benchmarks, surpassing the second-best model by 4.0 percent in accuracy on the general multi-oriented dataset.
Zhibin Ma, Pengwen Dai, Yi Liu +3
Shenzhen Campus of Sun Yat-sen University, China · Shenzhen Key Laboratory of Adversarial Artificial Intelligence, China · Baidu Inc., China +1
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.
Maximilian Woehrer
Research Group Software Architecture, Faculty of Computer Science, University of Vienna, Vienna, Austria · MAXI solutions e.U., Vienna, Austria
We present S2P-Net (Spectral-Spatial Polar Network), a compact deep learning architecture that achieves mathematically guaranteed rotation invariance without data augmentation. In this Paper, we also made a comparison to other neural network architectures (CNN`s). Have a look at the results and feel free to contact me for any questions. This is my first paper:) Made by Hackbert
Albert Heruth
Unaffiliated Researcher · Heide, Schleswig-Holstein, Germany · Schleswig-Holstein