Evaluating the Impact of Compression Techniques on the Robustness of CNNs under Natural Corruptions

Evaluating the Impact of Compression Techniques on the Robustness of CNNs under Natural Corruptions 1st Itallo Patrick Castro Alves Da Silva Computing Institute Federal University of Alagoas Macei´o, Brazil 0009-0008-8543-7776 2nd Emanuel Adler Medeiros Pereira Center of Technology Federal University of Rio Grande do Norte Natal, Brazil 0000-0002-6694-5336 3rd Erick de Andrade Barboza Computing Institute Federal University of Alagoas Macei´o, Brazil 0000-0002-0558-9120 4th Baldoino Fonseca dos Santos Neto Computing Institute Federal University of Alagoas Macei´o, Brazil 0000-0002-0730-0319 5th M´arcio de Medeiros Ribeiro Computing Institute Federal University of Alagoas Macei´o, Brazil 0000-0002-4293-4261 Abstract—Compressed deep learning models are crucial for deploying computer vision systems on resource-constrained de- vices. However, model compression may affect robustness, es- pecially under natural corruption. Therefore, it is important to consider robustness evaluation while validating computer vision systems. This paper presents a comprehensive evaluation of compression techniques—quantization, pruning, and weight clustering—applied individually and in combination to convolu- tional neural networks (ResNet-50, VGG-19, and MobileNetV2). Using the CIFAR-10-C and CIFAR-100-C datasets, we analyze the trade-offs between robustness, accuracy, and compression ratio. Our results show that certain compression strategies not only preserve but can also improve robustness, particularly on networks with more complex architectures. Utilizing multiobjec- tive assessment, we determine the best configurations, showing that customized technique combinations produce beneficial multi- objective results. This study provides insights into selecting compression methods for robust and efficient deployment of models in corrupted real-world environments. Index Terms—System Validation, Machine Learning, Image Classification, Robustness, Compression Techniques, Edge AI, TinyML. I. INTRODUCTION Humans can adapt to changes in image structures and styles, including snow, blur, and pixelation; however, computer vision models struggle with such variations. Consequently, the performance of the model decreases when the input is naturally distorted, which poses challenges in practical settings where such distortions are inevitable. For example, autonomous vehicles must handle diverse conditions such as fog, frost, snow, sandstorms, or falling leaves to accurately read traffic signs. However, predicting all natural conditions is not feasible. Therefore, evaluating the robustness of the model This study was financially supported by the National Council for Scientific and Technological Development (CNPq) project grant 404825/2023-0. is crucial to validate the reliability of computer vision and machine learning systems in safety-critical contexts [1]. The robustness of models against different types of per- turbation has been a studied topic in the machine learning community. Natural corruptions, which are an important type of disturbance, are common in real scenarios and can reduce the accuracy of models, so their study has been widely carried out [1]–[3]. Models are sometimes deployed on resource- limited devices, such as embedded systems and smartphones, necessitating a reduction in model size while maintaining ac- curacy. Techniques such as pruning (sparsity) [4], quantization [5], and weight sharing (clustering) [6] have been suggested for this purpose. These methods can be used individually or in combination to take advantage of their unique strengths in reducing the size of the model [7]. Therefore, it is important to study the robustness of these compressed models against natural corruptions, as they will be used in environments prone to corrupted images, exposing potential vulnerabilities. Works such as [8]–[11] applied com- pression techniques to machine learning models and evaluated the robustness of these optimized models. Although [11] ap- plies two or more successive techniques to reduce model, their study focuses on robustness against adversarial attacks, and the combinations of techniques explored were more limited. The purpose of this study is to analyze the impact of model compression techniques on robustness against natural corruption. From there, analyze the impact of these compres- sion techniques in relation to different models and also how robustness relates to other important metrics in relation to the compressed model. Our main contributions are as follows. • Evaluate the impact of compression techniques and their combinations on robustness under natural corruptions of models with different architectures. arXiv:2512.24971v1 [cs.CV] 31 Dec 2025 • Evaluate the trade-off between robustness, accuracy, and compression ratio. The structure of this paper is as follows: Section II surveys the literature pertinent to our research. Section III outlines the corruptions, compression methods, models, and evaluation criteria.

This content is AI-processed based on open access ArXiv data.