Next Article in Journal
An Efficient Approach for LBS Privacy Preservation in Mobile Social Networks
Previous Article in Journal
A Damage Model Reflecting the Interaction between Delamination and Intralaminar Crack for Failure Analysis of FRP Laminates
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Applied Sciences
Volume 9
Issue 2
10.3390/app9020315
Review Report
applsci-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
Article
Peer-Review Record

Chinese Character Boxes: Single Shot Detector Network for Chinese Character Detection

Appl. Sci. 2019, 9(2), 315; https://doi.org/10.3390/app9020315
by Junhwan Ryu and Sungho Kim *
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Appl. Sci. 2019, 9(2), 315; https://doi.org/10.3390/app9020315
Submission received: 22 December 2018 / Revised: 14 January 2019 / Accepted: 15 January 2019 / Published: 16 January 2019
(This article belongs to the Section Computing and Artificial Intelligence)

Round  1

Reviewer 1 Report

The paper is generally nicely laid out, and presented. Overall, well explained and clearly structured. There are some typos, grammatical and punctuation errors in the text, so, a proofread is necessary. Also, the conclusion section doesn’t add any new information to the paper, but rather it summarises the paper. Please re-write this section, and make sure there is some critical discussion added based on the results gained. Besides, the references must be checked for completeness.

Author Response

I will attach more details as a word file.

Author Response File: Author Response.doc

Reviewer 2 Report

The paper addresses detecting individual caoshu characters with a single-shot network and NMS.  The paper seems original and unique, as I haven’t seen much caoshu detection work.  However, more references to SSD papers would be a good idea.

The abstract could use some grammar checking/revision.  For example, "hand-craft features" should be "hand-crafted features".

When citing a paper such as "[20] used four networks..." I think it would be better to say "Maidana et. al [20] used four networks...".

NMS should be defined upon first usage in the paper, and also in the abstract.  

The last sentence of unit 2 should be more justified.  Why is it more important to accurately detect characters than sentences?

It's not clear why the 3 conv layers mentioned are 'unnecessary'.  

Equation 4 could use some definition of terms, GT is ground truth I think, but not sure exactly what FA stands for.

It’s not clear if “detection rate” is accuracy, or simply the percent of character detected.  An explanation would be helpful, and if it is accuracy, “accuracy” in parenthesis behind ‘detection rate’ would be useful.

How were the weights initialized?  This can change the configuration the net ends up in, and should be specificied.  Uniform, Glorot, He?  Ideally, the code with the network design and training setup would be published so that the research could be reproduced.

Author Response

I will attach more details as a word file.

Author Response File: Author Response.doc

Back to TopTop