Dead pixel test using effective receptive field
SCIE
SCOPUS
- Title
- Dead pixel test using effective receptive field
- Authors
- Kim, Bum Jun; Choi, Hyeyeon; Jang, Hyeonah; Lee, Dong Gu; Jeong, Wonseok; Kim, Sang Woo
- Date Issued
- 2023-03
- Publisher
- Elsevier BV
- Abstract
- Deep neural networks have been used in various fields, but their internal behavior in how they under- stand images is not well known. In this study, we discuss two counterintuitive properties of convolutional neural networks (CNNs). First, we evaluated the size of the receptive field of CNNs with their classification accuracy. Previous studies have attempted to increase the size of the receptive field for performance gain. However, we observed that some CNNs with a smaller receptive field can achieve higher classification accuracy. In this regard, we claim that a larger receptive field does not guarantee improved classification accuracy. Second, using the effective receptive field, we examined the contribution of each pixel to the output of CNN. Intuitively, each pixel is expected to equally contribute to the final output, but we found that there exist pixels in a partially dead state with little contribution to the output. We reveal that the reason for dead pixels lies in even stride operations with odd-sized kernels in CNN and propose a kernel padding method to remove the dead pixels. We demonstrated the vulnerability of CNNs with dead pix- els when we detect a noise or small box that is on dead pixels. Our findings on dead pixels should be understood and considered in practical applications of CNN.(c) 2023 Elsevier B.V. All rights reserved.
- URI
- https://oasis.postech.ac.kr/handle/2014.oak/119991
- DOI
- 10.1016/j.patrec.2023.02.018
- ISSN
- 0167-8655
- Article Type
- Article
- Citation
- Pattern Recognition Letters, vol. 167, page. 149 - 156, 2023-03
- Files in This Item:
- There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.