Learning by Watching via Keypoint Extraction and Imitation Learning

Round 1

Reviewer 1 Report

The abstract should be revised. The abstract was written by details focus on the idea of the paper. More than 50% from the abstract is motivation. Be specific and target your research contribution.  

• The Abstract to be enhanced, need additional technical contents.

• There is no story line in the introduction section, please add some motivations for this research in the introduction section.

• What are the different between your proposed method and other based techniques? 

• Figure 9 to 12 is NOT clear and need more discussions. 

• Please compare your proposed method with other based method in term of time complexity?

• Use some statistical significant test such as T-test or ANOVA to prove your improvements and quality of you method.

• List the limitations of your study in the conclusion section.

• The contributions of the paper is NOT clear, Sorry I must rejected.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This research proposes an architecture based on keypoint tracking and imitation learning as a two-fold method for learning from demonstrations. The idea of the paper is interesting, however, following suggestion are necessary for the improvement

1. Abstract is generic, author need to add paper contribution, motivation and results in this section briefly

2. Introduction paragraph (1) need citation

3. Add research motivation in the introduction part

4. add paper organization at the end of introduction part

5. better to mention paper contribution in bullets

6. Background section need improvement

7. References are not sufficient, add some latest references of 2021, and 2022 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

In this paper, the authors propose to learn by watching with keypoint extraction and imitation learning, Experimental results demonstrate the effectiveness of the proposed method. Some detailed comments can be found as follows:

1. The main contributions would be several points instead of only one point.

2. Why the performance of validation set is better than training set? Please clarify this point.

3. To provide better understanding for readers, some RL and IL-based image processing works are suggested to be reviewed, including RTN: Reinforced Transformer Network for Coronary CT Angiography Vessel-level Image Quality Assessment, Urban driving with conditional imitation learning, etc.

4. More state-of-the-art algorithms are suggested to be compared in the experiments.

5. It is suggested to further proofread the paper. For example, the summarized contribution is incomplete: “and then using image-to-image…” And many abbreviations lack full names.

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The author tried to modify the paper according to the comments I made earlier, but failed to achieve them, especially in comparisons and time complexity calculation. They also failed to find actual comparisons for statistical significant test. 

The technical contents is NOT well provided in the abstract section. The contributions of the paper still very weak. 

Author Response

We are highly grateful for your insightful comments on the manuscript. According to your suggestions, we have tried our best to revise most of the relevant parts of the manuscript.

Author Response File: Author Response.pdf

Reviewer 2 Report

I believe the manuscript has been sufficiently improved to warrant publication in Machines

Author Response

We are highly grateful for your insightful comments on the manuscript. According to your suggestions, we revised the relevant part of the manuscript.

Author Response File: Author Response.pdf

Reviewer 3 Report

The authors have addressed most of my comments. Can you give some literature that the performance of training set is better than the validation set? 

Author Response

We are highly grateful for your insightful comments on the manuscript. According to your suggestions, we revised the relevant part of the manuscript.

Author Response File: Author Response.pdf

Round 3

Reviewer 1 Report

The authors STILL NOT address my comments:

Please compare your proposed method with other based method in term of time complexity?

Use some statistical significant test such as T-test or ANOVA to prove your improvements and quality of you method.

The main contributions of the paper is NOT clear

The contributions of the paper still very weak.  

Author Response

please see Attachment

Author Response File: Author Response.pdf