Next Article in Journal
Monolingual and Cross-Lingual Intent Detection without Training Data in Target Languages
Next Article in Special Issue
Improving Heterogeneous Network Knowledge Transfer Based on the Principle of Generative Adversarial
Previous Article in Journal
Temperature Difference Sensor to Monitor the Temperature Difference in Processor Active Heat Sink Based on Thermopile
Previous Article in Special Issue
Crowd Counting Using End-to-End Semantic Image Segmentation
 
 
Article
Peer-Review Record

VRBagged-Net: Ensemble Based Deep Learning Model for Disaster Event Classification

Electronics 2021, 10(12), 1411; https://doi.org/10.3390/electronics10121411
by Muhammad Hanif *, Muhammad Atif Tahir and Muhammad Rafi
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Reviewer 3: Anonymous
Electronics 2021, 10(12), 1411; https://doi.org/10.3390/electronics10121411
Submission received: 30 April 2021 / Revised: 6 June 2021 / Accepted: 8 June 2021 / Published: 11 June 2021

Round 1

Reviewer 1 Report

The article “VRBagged-Net: Ensemble based Deep Learning Model for Disaster Event Classification” is very interesting and it is written well.

I only have two minor comments:

(1) The overall problem (disaster detection, etc.) can also be tackled using other media sources. These should be mentioned. As these approaches are out-of-scope, a suggest adding a single sentence / reference (e.g., “Improving the classification of flood tweets with contextual hydrological information in a multimodal neural network”, https://doi.org/10.1016/j.cageo.2020.104485)

(2) The proceedings of MediaEval 2020 (which have not been published yet) will contain your submission “An ensemble based method for the classification of flooding event using social media data”. Please describe the difference between your submitted article and your previous MediaEval 2020 contribution.

Author Response

Dear Reviewer 

 

Plz find attached our response for your feedback and comments.

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper focuses on leveraging deep learning techniques for flood classification. The authors utilized VGG and ResNet in the experiments. The main idea is based on Bootstrap aggregating (Bagging). Multiple networks are trained on datasets from stratified random sample selection. The outcomes of the networks are combined via majority voting for final prediction. The authors evaluated the proposed method on several datasets for flood classification. The results show improvement over existing methods. 


Although the novelty of the paper is limited, the method of the paper is clearly presented and reasonable. The results section is also easy to read. Since the authors leverage multiple networks in the training process, it is necessary to discuss the training time and memory consumption. 

There are several missing related works leveraging the deep neural networks for ensemble:


Deng, Li, and John C. Platt. "Ensemble deep learning for speech recognition." Fifteenth annual conference of the international speech 


Zhao, Yang, Jianping Li, and Lean Yu. "A deep learning ensemble approach for crude oil price forecasting." Energy Economics 66 (2017): 9-16.


Xiao, Yawen, et al. "A deep learning-based multi-model ensemble method for cancer prediction." Computer methods and programs in biomedicine 153 (2018): 1-9.


Guo, Yunhui, et al. "A broader study of cross-domain few-shot learning." European Conference on Computer Vision. Springer, Cham, 2020.

Author Response

Dear Reviewer 

Please find attached detail reply to your feedback and comments to our submission. 

Author Response File: Author Response.pdf

Reviewer 3 Report

This paper investigates an ensemble-based deep learning system for flood classification using social media. Various VGG and ResNet models are trained using multiple subsets of training data generated from the concept of bagging. The artical is nicely written, and the methodology is clear. The proposed model is evaluated on MediaEval tasks on flood detection. Interesting results are reported. However, there are some minor recommendations that need to be incorporated.

(a)    Section 1: Important review papers are needed to summarise (best) in a table format clearly indicating the performance achieved by various systems along with evaluation metrics used and its justification.

(b)    Section 2 describes the datasets for disaster response systems in detail. This section probably needs to move after Methodology.

(c)     Section 3.1.1: Authors discuss image resizing, but parameters are not described nor mentioned in this section.

(d)    Section 3.2.1: Both VGG and ResNet architectures need to be explained.

(e)    Section 4: Average precision needs to be mathematically defined.

(f)     Section 4.4: F1 score is quite low for all compared methods indicating the complexity of this dataset. Authors should give more insight about this task of MediaEval 2020.  

(g)    There are some references with missing journal / conference names e.g. [3], [6]. So you need to check all. 

Its a good paper, so I encourage its publication after the above revisions. 

Author Response

Dear Reviewer 

Please find attached detail response for the feedback and comments on our submitted paper.

thank you,

Author Response File: Author Response.pdf

Back to TopTop