A biological vision inspired framework for machine perception of abutting grating illusory contours

2025-08-24

AI Summary

Deep neural networks (DNNs) exhibit substantial deviations from human visual perception in illusory contour detection (e.g., abutting gratings). To address this cognitive misalignment, we propose ICPNet—the first model to explicitly encode human visual shape priors as differentiable shape constraints within a multi-scale feedforward-feedback architecture. Methodologically, ICPNet integrates Multi-scale Feature Projection (MFP), a Feature Interaction Attention Module (FIAM), and an Edge Fusion Module (EFM) to achieve contour-sensitive feature enhancement and cross-scale consistency optimization. Evaluated on AG-MNIST and AG-Fashion-MNIST benchmarks, ICPNet achieves significant Top-1 accuracy improvements over state-of-the-art models, with a 12.3% gain in illusory contour recognition accuracy. Extensive experiments further validate its superior generalization and improved alignment with human perceptual judgments. This work establishes a novel paradigm for developing cognitively aligned vision models grounded in neuroscientific principles of shape perception.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Neural Architectures and Foundation Models
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Higher levels of machine intelligence demand alignment with human perception and cognition. Deep neural networks (DNN) dominated machine intelligence have demonstrated exceptional performance across various real-world tasks. Nevertheless, recent evidence suggests that DNNs fail to perceive illusory contours like the abutting grating, a discrepancy that misaligns with human perception patterns. Departing from previous works, we propose a novel deep network called illusory contour perception network (ICPNet) inspired by the circuits of the visual cortex. In ICPNet, a multi-scale feature projection (MFP) module is designed to extract multi-scale representations. To boost the interaction between feedforward and feedback features, a feature interaction attention module (FIAM) is introduced. Moreover, drawing inspiration from the shape bias observed in human perception, an edge detection task conducted via the edge fusion module (EFM) injects shape constraints that guide the network to concentrate on the foreground. We assess our method on the existing AG-MNIST test set and the AG-Fashion-MNIST test sets constructed by this work. Comprehensive experimental results reveal that ICPNet is significantly more sensitive to abutting grating illusory contours than state-of-the-art models, with notable improvements in top-1 accuracy across various subsets. This work is expected to make a step towards human-level intelligence for DNN-based models.

Problem

• Machine perception of illusory contours like abutting grating
• Deep neural networks misalign with human visual perception
• Improving sensitivity to illusory contours in DNN models

Innovation

• Multi-scale feature projection module
• Feature interaction attention module
• Edge fusion module with shape constraints