You Always Recognize Me (YARM): Robust Texture Synthesis Against Multi-View Corruption

Weihang Ran, Wei Yuan, Yinqiang Zheng

Abstract: Damage to imaging systems and complex external environments often introduce corruption, which can impair the performance of deep learning models pretrained on high-quality image data. Previous methods have focused on restoring degraded images or fine-tuning models to adapt to out-of-distribution data. However, these approaches struggle with complex, unknown corruptions and often reduce model accuracy on high-quality data. Inspired by the use of warning colors and camouflage in the real world, we propose designing a robust appearance that can enhance model recognition of low-quality image data. Furthermore, we demonstrate that certain universal features in radiance fields can be applied across objects of the same class with different geometries. We also examine the impact of different proxy models on the transferability of robust appearances. Extensive experiments demonstrate the effectiveness of our proposed method, which outperforms existing image restoration and model fine-tuning approaches across different experimental settings, and retains effectiveness when transferred to models with different architectures. Code will be available at https://github.com/SilverRAN/YARM.

Lay Summary: Artificial intelligence technologies can learn visual features for classification using clean images and labels. However, in real-world application scenarios, the imaging system may capture low-quality images due to environmental factors, motion intensity, or hardware issues. We refer to such images as *corrupted images*. Previous studies have observed that severe corruption can significantly degrade the accuracy of AI models and have proposed various solutions to address this issue, such as restoring the corrupted images or fine-tuning models on datasets containing corrupted data. In contrast to these approaches, we draw inspiration from the camouflage patterns found in nature and propose to address the problem from the perspective of **data appearance**. Our proposed framework enables the redesign of an object’s original appearance, so that the images captured by the imaging system naturally exhibit resistance to corruption.

Link To Code: https://github.com/SilverRAN/YARM

Primary Area: Deep Learning->Robustness

Keywords: Image Corruption, 3D Reconstruction, NeRF

Submission Number: 11436