Review Reports

Round 1

Reviewer 1 Report

The authors propose a new lightweight Deep Convolutional Neural Network model for the binary classification of wildfire images and a new dataset transformation method that increases the number of dataset samples and improves the training and generalization of the deep learning model.

 

I recommend to correct some minor English errors (e.g. “which” instead of “that” on line 503 or “As a result…” instead of “As a results…” on line 516) and to review some writing aspects (e.g. line 445 is identic with line 446).

 

I recommend to indicate some references and/or give more details on some technical aspects as image tessellation (section 3.2.1, for example how “an image 4000 × 3000 could be downsized to 150 × 150 or 300 × 300 images”) or why real-time capabilities require 30 fps or 33.3 ms (line 373).

Author Response

We would like to thank reviewer #1 for his/her valuable comments on the manuscript. We have made the necessary modifications to our manuscript. Response to the reviewer’s comments is in the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

• It is appreciated that instead of using augmentation author has chosen tessellation.
• In 3.2.2 the F1 score is calculated for whole image or sub images. Also if ground truth is not present then how F1 can be calculated.
• Is the distribution of data for training and testing sub-images are generated from different image or from any combination (random).
• Please try with new image set and check model performance.
• As it is mentioned time required for prediction is sufficient to implement it real time. It can also be compared with other methods.

Author Response

We would like to thank reviewer #2 for his/her valuable comments on the manuscript. We have made the necessary modifications to our manuscript. Response to the reviewer’s comments is in the attachment.

Author Response File: Author Response.pdf