Artificial Visual System for Stereo-Orientation Recognition Based on Hubel-Wiesel Model

Abstract

Stereo-orientation selectivity is a fundamental neural mechanism in the brain that plays a crucial role in perception. However, due to the recognition process of high-dimensional spatial information commonly occurring in high-order cortex, we still know little about the mechanisms underlying stereo-orientation selectivity and lack a modeling strategy. A classical explanation for the mechanism of two-dimensional orientation selectivity within the primary visual cortex is based on the Hubel-Wiesel model, a cascading neural connection structure. The local-to-global information aggregation thought within the Hubel-Wiesel model not only contributed to neurophysiology but also inspired the development of computer vision fields. In this paper, we provide a clear and efficient conceptual understanding of stereo-orientation selectivity and propose a quantitative explanation for its generation based on the thought of local-to-global information aggregation within the Hubel-Wiesel model and develop an artificial visual system (AVS) for stereo-orientation recognition. Our approach involves modeling depth selective cells to receive depth information, simple stereo-orientation selective cells for combining distinct depth information inputs to generate various local stereo-orientation selectivity, and complex stereo-orientation selective cells responsible for integrating the same local information to generate global stereo-orientation selectivity. Simulation results demonstrate that our AVS is effective in stereo-orientation recognition and robust against spatial noise jitters. AVS achieved an overall over 90% accuracy on noise data in orientation recognition tasks, significantly outperforming deep models. In addition, the AVS contributes to enhancing deep models' performance, robustness, and stability in 3D object recognition tasks. Notably, AVS enhanced the TransNeXt model in improving its overall performance from 73.1% to 97.2% on the 3D-MNIST dataset and from 56.1% to 86.4% on the 3D-Fashion-MNIST dataset. Our explanation for the generation of stereo-orientation selectivity offers a reliable, explainable, and robust approach for extracting spatial features and provides a straightforward modeling method for neural computation research.

Keywords: Hubel-Wiesel model; artificial visual system; stereo-orientation selectivity.

Figure 1

The hierarchical cascade structure of the Hubel-Wiesel model.

Figure 2

The conceptually neural connections of stereo-orientation selectivity.

Figure 3

The information processing flow for local stereo-orientation information.

Figure 4

The remaining 12 types of stereo-orientation and their element indices within a local space.

Figure 5

The connection pattern between simple and complex stereo-orientation selective cell.

Figure 6

The overall process flow of AVS for stereo-orientation recognition.

Figure 7

One frame of drift grating stimuli data.

Figure 8

Activation records of complex stereo-orientation selective cells($0^{\circ }$, ${45}^{\circ }$, ${90}^{\circ }$, and ${135}^{\circ }$). The white area represents the rest state of stimuli, the gray area represents the moving state of stimuli, and the blue color denotes the neuron activation.

Figure 9

The generation process of oreintation information in random dots.

Figure 10

Acuuracy curve of AVS across different random dot datasets.

Figure 11

Instances of clean data and noise data.

Figure 12

The visualization of information processing within AVS.

Figure 13

The digital number ‘4’ and its composition of stereo-orientation information.

Figure 14

The separation of effective and negative information from original spatial information.

Figure 15

Ablation study on the feature selection process based on local stereo-orientation.

Figure 16

The point sampling effects with and without AVS processing.