ECDSA-Based Water Bodies Prediction from Satellite Images with UNet
, Samhitha Gadamsetty 1, Celestine Iwendi 2, Thippa Reddy Gadekallu 3,* and Imed Ben Dhaou 4,5,6
Round 1
Reviewer 1 Report
The work is very well presented and the results are clear and interested with a general scientific soundness. To make the processor more widely applicable with more general performance evaluation the paper would benefit of an analysis of the sensitiveness of the network to the quality of the images. Training has been performed with S2 images but it could be interesting to perform tests and present the results on different data sets i.e. with different GSD, SNR and if possible with band with differet spectral resolution.,
Author Response
1. To make the processor more widely applicable with more general performance evaluation the paper would benefit of an analysis of the sensitiveness of the network to the quality of the images.
2. Training has been performed with S2 images but it could be interesting to perform tests and present the results on different data sets i.e. with different GSD, SNR and if possible with band with different spectral resolution.,
Response:
As per your suggestion, a new dataset Landsat 8 OLI is used for comparison and stats are displayed in section 5.5
Reviewer 2 Report
I have major concerns regarding the methodology, discussion, grammatical errors, typos, and others which will be explained in the following comments:
- Spell-check is needed. For example, in L 270 and other sections, sentinetel-2 must be changed to sentinel-2.
- The studied area must be described. How much is the area? and etc.
- It is mentioned that there is a total of 10,000 images, but I think this is a very large number of images because the harvest period of Sentinel-2 is every 10 days since 2015.
- How did you deal with the cloud of images?
- Full names for all abbreviations should be presented in their first occurrences. For example, ECDSA in the abstract, and etc.
- Description of performance metrics must be transferred to the methodology.
Author Response
I have major concerns regarding the methodology, discussion, grammatical errors, typos, and others which will be explained in the following comments:
- Spell-check is needed. For example, in L 270 and other sections, sentinetel-2 must be changed to sentinel-2.
Response:
As per your suggestion, all the speck checks and typo errors are modified.
- The studied area must be described. How much is the area? and etc.
Response:
Section-2 literature survey gives a complete analysis of the studies focusing on UNet, Water bodies prediction and ECDSA.
- 3. It is mentioned that there is a total of 10,000 images, but I think this is a very large number of images because the harvest period of Sentinel-2 is every 10 days since 2015.
Response:
As per your suggestion, the image count is re-written. There is a total of 2841 sentinel-2 satellite images with corresponding 2841 masks. We have updated the same in abstract session.
- How did you deal with the cloud of images?
Response:
Once the dataset is given input to the model, the data pre-processing happens as described in the module 2 of the proposed methodology.
- Full names for all abbreviations should be presented in their first occurrences. For example, ECDSA in the abstract, and etc.
Response:
As per your suggestion the full forms for first occurrence of abbreviations are mentioned at respective places.
- Description of performance metrics must be transferred to the methodology.
Response:
As per your suggestion, performance metrics other than MeanIOU are shifted to Module 5 of Methodology on a high level.
Reviewer 3 Report
1. In Figures 1, 5, 6, 7,8,9,10,11，the annotations in the pictures are blurry.
2. The layout of the document is messy. Some paragraphs are indented with spaces before, and some are not. It is recommended to adjust the full text.
Author Response
Thank you for your positive feedback. We have updated the article as per the suggestions
Round 2
Reviewer 1 Report
I think that the additional test really shows the potentiality of the algorithm also on different data. I would suggest accepting the paper in the present form.
Reviewer 2 Report
The authors have considered all comments
