Next Article in Journal
Automated Schedule and Cost Control Using 3D Sensing Technologies
Next Article in Special Issue
Lightweight Micro-Expression Recognition on Composite Database
Previous Article in Journal
Blockchain Secured Dynamic Machine Learning Pipeline for Manufacturing
Previous Article in Special Issue
Depth-Adaptive Deep Neural Network Based on Learning Layer Relevance Weights
 
 
Article
Peer-Review Record

Unsupervised Semantic Segmentation Inpainting Network Using a Generative Adversarial Network with Preprocessing

Appl. Sci. 2023, 13(2), 781; https://doi.org/10.3390/app13020781
by Woo-Jin Ahn 1, Dong-Won Kim 2,*, Tae-Koo Kang 3, Dong-Sung Pae 4 and Myo-Taeg Lim 1,*
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Appl. Sci. 2023, 13(2), 781; https://doi.org/10.3390/app13020781
Submission received: 8 December 2022 / Revised: 26 December 2022 / Accepted: 30 December 2022 / Published: 5 January 2023
(This article belongs to the Special Issue Deep Learning Architectures for Computer Vision)

Round 1

Reviewer 1 Report

To solve the instability of generative adversarial network in semantically segmented image repair tasks, an unsupervised semantically segmented repair method using deep convolutional generative adversarial network is proposed, and new preprocessing methods and loss functions are introduced. The logic of the paper is clear and the method explanation is detailed, which clearly describes the overall framework of the network structure, the principle of the effectiveness of the algorithm. The results in the experimental part are clear and obvious, which  demonstrates the effectiveness of the algorithm. Some minor revisions maybe contribute to the improvement of the paper quality.

1. References are not uniformly formatted and the authors need to reorganize them.

2. It is better to polish your contributions in the Introduction part directly. 

3. The proposed SSIN is based on deep convolutional GAN (DCGAN), it is better to present the structure of DCGAN for clear description.

Author Response

Please refer to the attatched file. 

Author Response File: Author Response.pdf

Reviewer 2 Report

Title: Unsupervised Semantic Segmentation Inpainting Network Using a Generative Adversarial Network with Preprocessing

 1

 

keywords: Generative adversarial network; Adversarial Learning; Unsupervised learning; Inpainting; Semantic Segmentation. You have many keywords already in the title, it is not useful, keywords must be different.

2

Typo: Line 20: Fig ??

3

In Introduction, you use GAN twice, in page 1, line 26, 27, but it is explained only in page 3, line 76: „generative adversarial network (GAN)”.

4

Line 36: „segmentation amp” . Maybe „segmentation map”?

5

In Fig. 9 you have two Mercedes Logo. It is not the case, please remove it.

6

Line 185+4: „We introduce a new form of a total variation [31] loss function..”

But 31 is from 1992!

 

7.

After reading the entire work, a small contradiction arises in the mind of the reader. The introduction is written very well, defining and exemplifying the main operations that are the subject of the work. Figures 1, 2 and 3 exemplify the basic terminology, at the level of introduction to the field: Paired and Unpaired static and dynamic maps, even exemplification for Semantic Segmentation Inpainting. This approach indicates that the work is also addressed to beginners in the field. But as the research method is described, the scientific level is very high. In many paragraphs the need for additional explanations is felt.

 

Author Response

Please refer to the attatched file. 

Author Response File: Author Response.pdf

Back to TopTop