MU R-CNN: A Two-Dimensional Code Instance Segmentation Network Based on Deep Learning

Round 1

Reviewer 1 Report

Overall Idea is really much movel. 

However, need major revision before any final decision. 

Add all answers in revised version to strengthen yoru article. 

Comment file is attached.

Major Revision Needed:-

This paper is proposed to detect and location object services by considering object identification recognition based on vision sensors. They used several algorithms to make localization via modified IS and R-CNN. It is written very well, however, need following major revision and modifications very carefully. Add comments in revised version, necessarily

(In abstract section), Authors pinpoint the contributed work of their system. Revise again the abstract. (In Introduction section), Write first paragraph in which, authors should add few applications of object identification and tracking based on vision object analysis and processing areas. For example: tracking and detection [1-3], computer engineering [4,5], physical sciences, health-related issues [6], natural sciences and industrial academic areas [7]. And include all below reference.

[1] Human activity recognition from wearable sensors using extremely randomized trees,” in processing inter. conf. On electrical engineering and information comm. technology.  

[2] “A Triaxial acceleration-based human motion detection for ambient smart home system," IEEE International Conference on Applied Sciences and Technology, 2019.

[3] “Salient Segmentation based Object Detection and Recognition using Hybrid Genetic Transform”, IEEE ICAEM conference, 2019.

[4] “Advancements of image processing and vision in healthcare,” Journal of healthcare engineering, 2018.

[5] “Human body parts estimation and detection for physical sports movements,” IEEE International Conference on Communication, Computing and Digital Systems, 2019.

[6] “Students’ Behavior Mining in E-learning Environment Using Cognitive Processes with Information Technologies,” Education and Information Technologies, Springer, 2019.

[7] “Security architecture for third generation (3G) using GMHS cellular network,” in IEEE Conference on Emerging Technologies

(In Introduction section), In addition, authors must revise carefully the indoor and outdoor object identification scenarios based on vision cameras systems in 1 paragraphs. (In Introduction section), Authors should add information about different sensors used in image object capturing and processing such as binary, digital cameras and depth data in image analysis fields [1-7]. Also, give following references for different sensor based technologies.

[1] A real-time system for object detection and location reminding with RGB-D camera, in proc. conf. on consumer electronics, 2014.

[2] “Multi-features descriptors for human activity tracking and recognition in Indoor-outdoor environments,” IEEE International Conference on Applied Sciences and Technology, 2019.

[3] Digital image processing in remote sensing,” in proc. conf. on computer graphics and image processing, 2009.

[4] Ridge body parts features for human pose estimation and recognition from RGB-D video data,” in Proceedings of the IEEE International Conference on computing, communication and networking technologies.

[5] “Wearable Sensor-Based Human Behavior Understanding and Recognition in Daily Life for Smart Environments, IEEE conference on International Conference on Frontiers of information technology, 2018.

[6] “Robust spatio-temporal features for human interaction recognition via artificial neural network, IEEE conference on International Conference on Frontiers of information technology, 2018.

[7] Satellite image processing and air pollution detection,” in Proceedings of the IEEE International Conference on Acoustics, speech and signal processing.

(In Introduction section), 2^nd, and 3^rd paragraphs need to be merged as one. (In Related work section), Make 2^nd paragraph in which; Mentioned about the state of the art features used in different areas of object tracking and understanding fields and mentioned below references [1-7].

[1] “Real-time continuous feature extraction in large size satellite images,” Journal of systems architecture: the EUROMICRO, 2016.

[2] “Depth maps-based human segmentation and action recognition using full-body plus body color cues via recognizer engine, Journal of Electrical Engineering & Technology, 2018.

[3] “Facial Expression Recognition in Image Sequences Using 1D Transform and Gabor Wavelet Transform, IEEE conference on International Conference on Applied and Engineering Mathematics, 2018.

[4] Vision-based real-time motion capture system using multiple cameras,” in Proceedings IEEE Conference on Multisensor Fusion and Integration for Intelligent Systems.

[5] "Detecting Complex 3D Human Motions with Body Model Low-Rank Representation for Real-Time Smart Activity Monitoring System," KSII Transactions on Internet and Information Systems, vol. 12, no. 3, pp. 1189-1204, 2018.

[6] “Facial expression recognition using hybrid features and self-organizing maps,” in Proceedings IEEE International Conference on multimedia and expo, Jul. 2017.

[7] Person re-identification across multi-camera system based on local descriptors,” in Proceedings IEEE conference on distributed smart cameras.

(In Related work section), Remove old references as 9, 17, 21, 22, 23, 24, 25, 31, 33, 35, 36, 37, 39. (In Related work section), Object tracking methods are not well introduced. Add few lines (In Our contribution section), Add Figure in which system architecture is defined. (In Our contribution section), Pre-processing step is totally ignored. Please mentioned about algorithms and techniques used to performed de-noising, detection, segmentations and so on. (In Our contribution section), Authors need to add different object tracking techniques for scene recognition and add below references to strengthen this section.

[1] Gaussian mixture based HMM for human daily activity recognition using 3D skeleton features,” in processing inter. conf. On Industrial electronics and applications. 

[2] “A depth video-based human detection and activity recognition using multi-features and embedded hidden Markov models for health care monitoring systems, International Journal of Interactive multimedia and Artificial Intelligence, vol. 4(4), pp. 54-62, 2017.

[3] “Robust human activity recognition from depth video using spatiotemporal multi-fused features, Pattern recognition, 2017

[4] “Individual Detection-Tracking-Recognition using depth activity images,” in Proceedings 12th IEEE International Conference on Ubiquitous Robots and Ambient Intelligence, pp. 450-455, 2015.

[5] “Depth Images-based Human Detection, Tracking and Activity Recognition Using Spatiotemporal Features and Modified HMM, Journal of Electrical Engineering and Technology, pp. 1921-1926, 2016.

[6] “Human depth sensors-based activity recognition using spatiotemporal features and hidden markov model for smart environments, Journal of computer networks and communications, vol. 2016, pp. 1-11, 2016.

[7] “Facial Expression recognition using 1D transform features and Hidden Markov Model, Journal of Electrical Engineering & Technology, vol. 12(4), pp. 1657-1662, 2017.

[8] “Human activity recognition based on the combined SVM & HMM,” in Proceedings Inter. Conf. on Information and Automation.

(In Our contribution), this section is conventional work. Known to all reader. Please reduce. (In Our contribution), In Figure 5, sub-function are ignored especially in RPN and UNet? Add sub-blocks. (In Our contribution), How authors are concise different features values at same feature vector? (In Our contribution), In Figure 6, why authors are restricted with conv 3x3 and pllo 2x2. Show us some statistical Table to defend this matrix? (Experiment Results), First; write datasets descriptions clearly. How many datasets, images for training and testing. Please provides few other examples of datasets. Not just one. (Experiment Results), Add a new Table and calculate performance values between state of the art systems with your system. Use these state of the art systems. i.e., HMM, Modified HMM, embedded HMM, GMM, SVMs, etc. [1-6]. Authors must refer following references.

[1] Gaussian mixture based HMM for human daily activity recognition using 3D skeleton features,” in processing inter. conf. On Industrial electronics and applications.  

[2] “A hybrid feature extraction approach for human detection, tracking and activity recognition using depth sensors, Arabian Journal for Science and Engineering, vol. 41(3), pp. 1043-1051, 2016.

[3] “Depth Silhouettes Context: A new robust feature for human tracking and activity recognition based on embedded HMMs,” in Proceedings 12th IEEE International Conference on Ubiquitous Robots and Ambient Intelligence, pp. 294-299, 2015.

[4] “Human activity recognition based on the combined SVM & HMM,” in Proceedings Inter. Conf. on Information and Automation.

[5] “Dense RGB-D Map-Based Human Tracking and Activity Recognition using Skin Joints Features and Self-Organizing Map ,” KSII Transactions on internet and information systems, vol. 9(5), pp. 1856-1869, 2015.

[6] A spatiotemporal motion variation features extraction approach for human tracking and pose-based action recognition,” in Proceedings IEEE International Conference on Informatics, electronics and vision, 2015.

(Conclusion and Future work), Authors must have mentioned about failures and weak areas of their system.

Comments for author File: Comments.pdf

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 2 Report

The authors study the problem of segmenting instances of QR codes in images. Such task requires high precision in locating the QR code, in order to properly decode it afterward. The solution presented is based on convolutional neural networks and combines an instance of the popular U-Net with a region proposal network.

 

The overall methodology looks sound and is thoroughly explained. This is more of an application paper, since it illustrates a useful technique to solve the problem, yet it does not provide a major theoretical contribution.

 

The paper is quite wordy and spends quite a lot of paragraphs explaining the background of CNNs and the method proposed itself. Readers who are familiar with CNNs may find it hard to properly identify the contribution of the paper, concisely, since it gets lost in all the other more elementary explanations.

 

Other than that, I don’t have major comments.

 

I do suggest to revise the use of English language, since there are lots of minor mistakes here and there, especially in the use of articles. E.g., a sentence states “the following experiments is carried out”.

 

Since I am not very familiar with this application domain, I cannot much judge the appropriateness of the experimental comparisons agains other methods in the field.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 3 Report

The paper is interesting and has novelty. It is well written and has a good strucute and authors present their work in a consice and compreheisive way. I really liked it and i found it to be sound. However, i feel that in the introduction authors should highlight the controibutions more explicitly as well as the way that it could be assistive to the realted research community. The resutls are indeed interesting , however a better connection and discussion to realted works should be made.

 

All in all the paper is interesting it has novelty and contribution and i vote for its acceptance.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

All questions are revised very well. 

definitely, accepted. 

Congratulation.