Visual Localization and Target Perception Based on Panoptic Segmentation
Round 1
Reviewer 1 Report
This paper presents a visual localization technique using semantic information. The proposed technique integrates panoptic segmentation and the matching network to refine the sensor’ position and orientation, and complete the target perception.
The manuscript is well written and organized and the reported results and comparisons are convincing.
I recommend to add a pseudocode of the proposed methodology and correct some minor typos in the text.
Finally, an evaluation of the time costs for the proposed methodology must be performed.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
An improved algorithm is presented, which seems to make a long-term visual localisation more accurate in a variety of environments, although computational cost is too high for real-time applications. Anyways, the algorithm may help as a ground truth source when faster NN solutions are being developed.
There are few typos, which could be corrected:
P3: *Extensive --> Extensive
P6: posse --> pose
P14: results.by --> results. By
P15: peception --> perception
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The authors partially replied to my comments and suggestions.
However, what has been done during this review round is enough to make this manuscript publishable in its present form. So I can advise to accept it.
