Evaluation of Deep Learning-Based Segmentation Methods for Industrial Burner Flames

Round 1

Reviewer 1 Report

Most of the questions have been answered by the author's answer, so I will deliver an accept opinion.

Reviewer 2 Report

The manuscript presents interesting and important results and the the research methodology was explained precisely and clearly. The work presented by the authors is original and innovative. The research methodology was explained precisely and clearly. The results obtained suggest that the method proposed may be interesting to researchers working on combustion process control but also for those interested in applying deep learning based methods to image segmentation in broader applications. In my opinion, the manuscript is well written and should be of great interest to the readers. I recommend accepting the manuscript for publication in its present form.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

1. This study targets subjects and images that can only be applied in special circumstances, and it is judged that the research method is not applicable to other environmental conditions. In addition, since the research method applied to this study is already known, it seems that there is not a great deal of benefit or benefit to other researchers through publish.

3. DeepLab v3 is known as the semantic segmentation model, and Mask RCNN is known as the instance segmentation model. There is not enough explanation for the reason that the problem that can be solved by instance segmentation is implemented with the semantic segmentation model, and it is judged that it is difficult to find similar previous studies.

3. Researchers use DeepLab v3 on various backbone networks such as ResNet18 and ResNet50.
Although the applied architecture is presented, the research credibility is low because there is no explanation for the diagram for the architecture.

4. The researcher said that transfer learning was applied, but there is no mention of the degree of fine-tuning, so an important explanation of the research method is omitted.

Reviewer 2 Report

The manuscript presents interesting and important results and the the research methodology was explained precisely and clearly. I recommend accepting the manuscript for publication in its present form.

Reviewer 3 Report

The authors discuss the current topic of the ability to access deep learning transient segmentation of the flames of industrial burners from a waste incinerator. this topic excellently fulfils the focus of the journal Energies and the used scientific methods are appropriately selected. The possibilities of further research are discussed in the paper as well. The conclusion is supported by the presented results. The article might be published in the present form.

best regards

Reviewer