Mimicking Human Visual Development for Learning Robust Image Representations

2025-12-16

AI Summary

Convolutional neural networks (CNNs) exhibit limited out-of-distribution generalization and noise robustness compared to the human visual system. Method: Inspired by infant visual development, we propose a learnable curriculum that progressively sharpens input images during training—starting from highly blurred inputs and gradually increasing clarity—to prioritize global structural learning and suppress high-frequency noise. Contrary to the conventional belief that early-stage blurring harms performance, our approach enhances robustness without sacrificing in-distribution accuracy. Technically, it integrates Gaussian blur scheduling within a curriculum learning framework and is compatible with augmentation strategies such as CutMix and MixUp, jointly optimizing natural and adversarial robustness. Results: On CIFAR-10-C and ImageNet-100-C, our method reduces mean Corruption Error (mCE) by 8.30% and 4.43%, respectively, and demonstrates superior robustness against common corruptions, distortions, and PGD attacks.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

The human visual system is remarkably adept at adapting to changes in the input distribution; a capability modern convolutional neural networks (CNNs) still struggle to match. Drawing inspiration from the developmental trajectory of human vision, we propose a progressive blurring curriculum to improve the generalization and robustness of CNNs. Human infants are born with poor visual acuity, gradually refining their ability to perceive fine details. Mimicking this process, we begin training CNNs on highly blurred images during the initial epochs and progressively reduce the blur as training advances. This approach encourages the network to prioritize global structures over high-frequency artifacts, improving robustness against distribution shifts and noisy inputs. Challenging prior claims that blurring in the initial training epochs imposes a stimulus deficit and irreversibly harms model performance, we reveal that early-stage blurring enhances generalization with minimal impact on in-domain accuracy. Our experiments demonstrate that the proposed curriculum reduces mean corruption error (mCE) by up to 8.30% on CIFAR-10-C and 4.43% on ImageNet-100-C datasets, compared to standard training without blurring. Unlike static blur-based augmentation, which applies blurred images randomly throughout training, our method follows a structured progression, yielding consistent gains across various datasets. Furthermore, our approach complements other augmentation techniques, such as CutMix and MixUp, and enhances both natural and adversarial robustness against common attack methods. Code is available at https://github.com/rajankita/Visual_Acuity_Curriculum.

Problem

• Improves CNN generalization and robustness via progressive blurring curriculum.
• Reduces corruption error on datasets like CIFAR-10-C and ImageNet-100-C.
• Enhances robustness against distribution shifts and noisy inputs.

Innovation

• Progressive blurring curriculum mimics human visual development
• Training starts with blurred images, gradually reduces blur
• Enhances generalization and robustness against distribution shifts