Dendritic Convolution for Noise Image Recognition

TL;DR

This work tackles the challenge of noise robustness in image recognition by proposing Dendritic Convolution (DDC), a plug in replacement for standard convolution that emulates nonlinear dendritic integration. DD C introduces a local neighborhood interaction term via a Hadamard like mechanism and a residual dendritic structure that preserves trunk information while injecting nonlinear branches, enabling adaptive noise filtering at feature extraction. Empirical results on CIFAR-10 and VOC2007 across multiple backbones show consistent anti noise gains, with up to 19.8% mAP improvement in detection and notable accuracy boosts in classification, especially for lightweight models. The approach offers a biologically inspired, bottom layer enhancement that improves robustness without additional modules or data preprocessing, suggesting practical impact for real world noisy visual systems.

Abstract

In real-world scenarios of image recognition, there exists substantial noise interference. Existing works primarily focus on methods such as adjusting networks or training strategies to address noisy image recognition, and the anti-noise performance has reached a bottleneck. However, little is known about the exploration of anti-interference solutions from a neuronal perspective.This paper proposes an anti-noise neuronal convolution. This convolution mimics the dendritic structure of neurons, integrates the neighborhood interaction computation logic of dendrites into the underlying design of convolutional operations, and simulates the XOR logic preprocessing function of biological dendrites through nonlinear interactions between input features, thereby fundamentally reconstructing the mathematical paradigm of feature extraction. Unlike traditional convolution where noise directly interferes with feature extraction and exerts a significant impact, DDC mitigates the influence of noise by focusing on the interaction of neighborhood information. Experimental results demonstrate that in image classification tasks (using YOLOv11-cls, VGG16, and EfficientNet-B0) and object detection tasks (using YOLOv11, YOLOv8, and YOLOv5), after replacing traditional convolution with the dendritic convolution, the accuracy of the EfficientNet-B0 model on noisy datasets is relatively improved by 11.23%, and the mean Average Precision (mAP) of YOLOv8 is increased by 19.80%. The consistency between the computation method of this convolution and the dendrites of biological neurons enables it to perform significantly better than traditional convolution in complex noisy environments.

Figures (8)

• Figure 1: Schematic Diagram of Neuron Simulation Comparison Between Traditional Convolution and Dendritic Convolution
• Figure 2: Schematic diagram of Traditional Convolution operation.
• Figure 3: Schematic diagram of DDC feature extraction.
• Figure 4: Visualization of Noisy Images; Figure (a) shows samples from the classification noise dataset of CIFAR-10, and Figure (b) shows samples from the detection noise dataset of VOC2007.
• Figure 5: Accuracy comparison of classification models under different noises. The abscissa represents the model type, and the ordinate represents the classification accuracy (%). The yellow bar graph represents the performance of the model using traditional convolution ( Conv ), and the green bar graph represents the performance of the model using DDC. All data are the average of three test results.