Computer graphics identification combining convolutional and recurrent neural networks
In this letter, a deep-learning-based pipeline is proposed to distinguish photographics (PGs) from computer-graphics (CGs) combining convolutional neural network (CNN) and recurrent neural network (RNN). In the preprocessing stage, the color space transformation and the Schmid filter bank are utiliz...
Saved in:
Main Authors: | , , , |
---|---|
Other Authors: | |
Format: | Article |
Language: | English |
Published: |
2020
|
Subjects: | |
Online Access: | https://hdl.handle.net/10356/142572 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
id |
sg-ntu-dr.10356-142572 |
---|---|
record_format |
dspace |
spelling |
sg-ntu-dr.10356-1425722020-06-24T08:05:51Z Computer graphics identification combining convolutional and recurrent neural networks He, Peisong Jiang, Xinghao Sun, Tanfeng Li, Haoliang School of Electrical and Electronic Engineering Engineering::Electrical and electronic engineering Computer-graphics (CGs) Convolutional Neural Network (CNN) In this letter, a deep-learning-based pipeline is proposed to distinguish photographics (PGs) from computer-graphics (CGs) combining convolutional neural network (CNN) and recurrent neural network (RNN). In the preprocessing stage, the color space transformation and the Schmid filter bank are utilized to extract chrominance and luminance components, which suppress the irrelevant information of various image contents for the CG identification task. Then, a dual-path CNN architecture is designed to learn joint feature representations of local patches for exploiting their color and texture characteristics. To extract the global artifact, the directed acyclic graph RNN is applied to model the spatial dependence of local patterns. Finally, the output score of RNN is used to identify the input sample. The CG/PG dataset is constructed by collecting samples from the Internet. Experimental results show that the proposed framework can outperform state-of-The-Art methods on identification ability of CGs, especially for images with low resolution. 2020-06-24T08:05:51Z 2020-06-24T08:05:51Z 2018 Journal Article He, P., Jiang, X., Sun, T., & Li, H. (2018). Computer graphics identification combining convolutional and recurrent neural networks. IEEE Signal Processing Letters, 25(9), 1369-1373. doi:10.1109/LSP.2018.2855566 1070-9908 https://hdl.handle.net/10356/142572 10.1109/LSP.2018.2855566 2-s2.0-85049927259 9 25 1369 1373 en IEEE Signal Processing Letters © 2018 IEEE. All rights reserved. |
institution |
Nanyang Technological University |
building |
NTU Library |
country |
Singapore |
collection |
DR-NTU |
language |
English |
topic |
Engineering::Electrical and electronic engineering Computer-graphics (CGs) Convolutional Neural Network (CNN) |
spellingShingle |
Engineering::Electrical and electronic engineering Computer-graphics (CGs) Convolutional Neural Network (CNN) He, Peisong Jiang, Xinghao Sun, Tanfeng Li, Haoliang Computer graphics identification combining convolutional and recurrent neural networks |
description |
In this letter, a deep-learning-based pipeline is proposed to distinguish photographics (PGs) from computer-graphics (CGs) combining convolutional neural network (CNN) and recurrent neural network (RNN). In the preprocessing stage, the color space transformation and the Schmid filter bank are utilized to extract chrominance and luminance components, which suppress the irrelevant information of various image contents for the CG identification task. Then, a dual-path CNN architecture is designed to learn joint feature representations of local patches for exploiting their color and texture characteristics. To extract the global artifact, the directed acyclic graph RNN is applied to model the spatial dependence of local patterns. Finally, the output score of RNN is used to identify the input sample. The CG/PG dataset is constructed by collecting samples from the Internet. Experimental results show that the proposed framework can outperform state-of-The-Art methods on identification ability of CGs, especially for images with low resolution. |
author2 |
School of Electrical and Electronic Engineering |
author_facet |
School of Electrical and Electronic Engineering He, Peisong Jiang, Xinghao Sun, Tanfeng Li, Haoliang |
format |
Article |
author |
He, Peisong Jiang, Xinghao Sun, Tanfeng Li, Haoliang |
author_sort |
He, Peisong |
title |
Computer graphics identification combining convolutional and recurrent neural networks |
title_short |
Computer graphics identification combining convolutional and recurrent neural networks |
title_full |
Computer graphics identification combining convolutional and recurrent neural networks |
title_fullStr |
Computer graphics identification combining convolutional and recurrent neural networks |
title_full_unstemmed |
Computer graphics identification combining convolutional and recurrent neural networks |
title_sort |
computer graphics identification combining convolutional and recurrent neural networks |
publishDate |
2020 |
url |
https://hdl.handle.net/10356/142572 |
_version_ |
1681056256844890112 |