Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Effects of Gabor Filters on Classification Performance of CNNs Trained on a Limited Number of Conditions

Akito Morita, Hirotsugu Okuno

TL;DR

A Gabor filter, a model of the feature extractor of the VNS, is used as a preprocessor for CNNs to investigate the accuracy of the CNNs trained with small amounts of data and showed that preprocessing with Gabor filters improves the generalization performance of CNNs and contributes to reducing the size of CNNs.

Abstract

In this study, we propose a technique to improve the accuracy and reduce the size of convolutional neural networks (CNNs) running on edge devices for real-world robot vision applications. CNNs running on edge devices must have a small architecture, and CNNs for robot vision applications involving on-site object recognition must be able to be trained efficiently to identify specific visual targets from data obtained under a limited variation of conditions. The visual nervous system (VNS) is a good example that meets the above requirements because it learns from few visual experiences. Therefore, we used a Gabor filter, a model of the feature extractor of the VNS, as a preprocessor for CNNs to investigate the accuracy of the CNNs trained with small amounts of data. To evaluate how well CNNs trained on image data acquired under a limited variation of conditions generalize to data acquired under other conditions, we created an image dataset consisting of images acquired from different camera positions, and investigated the accuracy of the CNNs that trained using images acquired at a certain distance. The results were compared after training on multiple CNN architectures with and without Gabor filters as preprocessing. The results showed that preprocessing with Gabor filters improves the generalization performance of CNNs and contributes to reducing the size of CNNs.

Effects of Gabor Filters on Classification Performance of CNNs Trained on a Limited Number of Conditions

TL;DR

A Gabor filter, a model of the feature extractor of the VNS, is used as a preprocessor for CNNs to investigate the accuracy of the CNNs trained with small amounts of data and showed that preprocessing with Gabor filters improves the generalization performance of CNNs and contributes to reducing the size of CNNs.

Abstract

In this study, we propose a technique to improve the accuracy and reduce the size of convolutional neural networks (CNNs) running on edge devices for real-world robot vision applications. CNNs running on edge devices must have a small architecture, and CNNs for robot vision applications involving on-site object recognition must be able to be trained efficiently to identify specific visual targets from data obtained under a limited variation of conditions. The visual nervous system (VNS) is a good example that meets the above requirements because it learns from few visual experiences. Therefore, we used a Gabor filter, a model of the feature extractor of the VNS, as a preprocessor for CNNs to investigate the accuracy of the CNNs trained with small amounts of data. To evaluate how well CNNs trained on image data acquired under a limited variation of conditions generalize to data acquired under other conditions, we created an image dataset consisting of images acquired from different camera positions, and investigated the accuracy of the CNNs that trained using images acquired at a certain distance. The results were compared after training on multiple CNN architectures with and without Gabor filters as preprocessing. The results showed that preprocessing with Gabor filters improves the generalization performance of CNNs and contributes to reducing the size of CNNs.
Paper Structure (13 sections, 1 equation, 4 figures, 4 tables)

This paper contains 13 sections, 1 equation, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Image Classification Algorithm
  3. Processing Flow
  4. Gabor Filters
  5. Dataset
  6. Experimental Setups
  7. Evaluation Method
  8. CNN Architecture and Parameters
  9. Data Augmentations
  10. Results
  11. Comparison of Accuracy
  12. Evaluation of Feature Vectors Represented by Each Layer
  13. Conclusion

Figures (4)

  • Figure 1: Processing flow. (a) Processing flow without Gabor filters. (b)-(d) Processing flow with Gabor filters. The rectangles below Gabor filters represent rectification.
  • Figure 2: Examples of Gabor filtered images. (a) Input image. (b)-(e) Gabor filtered images. (b), (c), (d), and (e) are results for $\theta = 0, \pi/4, \pi/2, 3\pi/4$, respectively.
  • Figure 3: Environment for image acquisition and example images of the dataset. (a) Environment for image acquisition. (b) The frame used for camera height adjustment. (c) Examples of images taken from five different heights using the frame shown in (b). The heights of the camera positions are 34.0, 28.0, 22.0, 16.0, and 10.0 cm from the left image. (d) Example images of 10 objects in the dataset.
  • Figure 4: Classification accuracy of linear SVMs that use feature vectors generated by each residual block of the ResNet18 as input. The layer number indicates the depth of the layer, where 1 indicates the first residual block and 8 indicates the last residual block.