Automatic Weight Prediction System for Korean Cattle Using Bayesian Ridge Algorithm on RGB-D Image

Round 1

Reviewer 1 Report

This paper proposed an automatic weight prediction system for Korean cattle using RGB-D images and various computer vision techniques. Different regression models were applied and evaluated. Overall the study is interesting and has good results.

Page 2: "In addition, this process may give a lot of stress to the livestock, which may reduce the weight of the livestock." - Please support with evidence/reference this sentence.

Table 1 and Table 2 - It's not clear that what does Total (Male + Female) mean here? How are total and individual values related to each other? Why are the individual values of the coefficient of determination less than the total?

Author Response

We attached the response letter.

Author Response File: Author Response.pdf

Reviewer 2 Report

Proposed manuscript deals with a very interesting topic of automatic contactless weight prediction based on RGB-D camera. The topic is interestng from both the infromation engineering point of view (depth sensor, image processing, machine learning) and the livestock raising point of view (an area where modern methods of image analysis are increasingly used).

This paper introduce a new methodology of image processing (of Korean cattle) bssed on (i) image segmentation, (ii) feature extraction and (iii) prediction models, where methods of machine learning are employed.

I have following minor commens:

- the method is based on images of 15 pieces of Korean cattle, I think it should be much more
- whould it be possibe to increase the precision using fusion from more sensors?

Owing fact mentioned above, I recommend to accept the manuscript after MINOR REVISION.

Author Response

We attached the response letter.

Author Response File: Author Response.pdf

Reviewer 3 Report

The subject treated seems interesting but the methods used are not new. In general, the manuscript is well written although it needs to be proofread in terms of English.

The case of pre-processing, for example when an image is noisy, has not been considered. However, it is important in real life to consider a simple but efficient denoising procedure for example by wavelets or directly during the acquisition by the concept of compressed sensing. In this case, the following references should be considered:

A review of wavelet denoising in medical imaging. In Proceedings of the International Workshop on Systems, Signal Processing and Their Applications (IEEE/WOSSPA’13), Algiers, Algeria, 12–15 May 2013; pp. 19–26

Image Denoising Using a Compressive Sensing Approach Based on Regularization Constraints. Sensors 2022, 22, 2199. https://doi.org/10.3390/s22062199

It is useful to analyze works close to yours:

Contour Extraction of Individual Cattle From an Image Using Enhanced Mask R-CNN Instance Segmentation Method.  IEEE Access, vol. 9, pp. 56984-57000, 2021, doi: 10.1109/ACCESS.2021.3072636.

The computer vision aspect is a promising approach and it is also important to cite the following key references in this context:

Past, Present, and Future of Face Recognition: A Review. Electronics 2020, 9, 1188.  https://doi.org/10.3390/electronics9081188

Ear Recognition Based on Deep Unsupervised Active Learning. IEEE Sens. J. 2021, 21, 20704–20713. , doi: 10.1109/JSEN.2021.3100151

Finally, today the use of deep learning is to be considered as it is the case of the following key references:

An Intelligent Pig Weights Estimate Method Based on Deep Learning in Sow Stall Environments.  IEEE Access, vol. 7, pp. 164867-164875, 2019, doi: 10.1109/ACCESS.2019.2953099

Moreover, the manuscript has some shortcomings as for example the table 19 summarizing the computational complexity is not based on any justification and should be deleted.  You recall an ANN structure (Figure 12) without using it whereas it is important to exploit this way and to display the corresponding metrics (Accuracy, Precision, Recall...).

Finally, the references must be written according to the MDPI standard.

Author Response

We attached the response letter.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

The manuscript has been improved by taking into account the reviewers' recommendations.
It remains to write correctly the references. Indeed, the authors must put the name of each author followed by the initial of his first name and not the opposite. Moreover, in reference 9, the authors are not mentioned.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf