Gabor Convolutional Networks

By Shangzhen Luan, Baochang Zhang, Jungong Han, Jianzhuang Liu. IEEE Workshop/Winter Conference on Applications of Computer Vision, 2017

395 citations recorded by the source corpus.

Steerable properties dominate the design of traditional filters, e.g., Gabor filters, and endow features the capability of dealing with spatial transformations. However, such excellent properties have not been well explored in the popular deep convolutional neural networks (DCNNs). In this paper, we propose a new deep model, termed Gabor Convolutional Networks (GCNs or Gabor CNNs), which incorporates Gabor filters into DCNNs to enhance the resistance of deep learned features to the orientation and scale changes. By only manipulating the basic element of DCNNs based on Gabor filters, i.e., the convolution operator, GCNs can be easily implemented and are compatible with any popular deep learning architecture. Experimental results demonstrate the super capability of our algorithm in recognizing objects, where the scale and rotation changes occur frequently. The proposed GCNs have much fewer learnable network parameters, and thus is easier to train with an endtoend pipeline. The source code will be here 1.

Subfields: Computer Vision and Pattern Recognition, Media Technology

View Paper: https://www.semanticscholar.org/paper/6993df15fe72285f203ac09679854d96e208c0d5

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Guoying Zhao
2. Li Liu
3. Qi Tian
4. Matti Pietikäinen
5. Xiaogang Wang
6. Feng Zhou
7. Xiaobai Li
8. Yi Yang

In Media Technology

1. Qi Tian
2. Matti Pietikäinen
3. Guoying Zhao
4. Alberto Signoroni
5. Zitong Yu
6. Zhuo Su
7. Feng Zhou
8. Annalisa Baronio

Citations in the corpus

• Pixel Difference Networks for Efficient Edge Detection (2021, methodology): “On the other hand, Gabor convolution [34] encodes the orientation and scale information in the convolution kernels by multiplying the kernels with a group of Gabor filters, while PDC is more compact without any auxiliary traditional feature filters.”
• Searching Central Difference Convolutional Networks for Face Anti-Spoofing (2020, background): “Here we discuss the relations between CDC and vanilla convolution, local binary convolution[27] and gabor convolution[38], which share”
• Deep Learning Meets Hyperspectral Image Analysis: A Multidisciplinary Review (2019, methodology): “PCA was used also in [103] to pre-process medical HSI data and improved performance was obtained by the combination or modulation of CNN kernels with Gabor kernels in the preliminary network layers, as suggested in [165].”