# Critique of papers
# Introduction
In the introduction, cite the journal article in full and provide a summary of the journal article. Provide the background to this piece of research, establishing its place within the field. Use the answers to the questions in Establishing the Research Context to develop this section. Then use the answers to the questions in the section Analyzing the Text to further develop the section.
# Establishing the Research Context
Once you are reasonably familiar with the article, it is important to gain an understanding of the RESEARCH CONTEXT, both societal and intellectual. To establish the research context, questions such as the following should be addressed:
- Who conducted the research? What were/are their interests?
- When and where was the research conducted?
- Why did they do this research?
- Was this research pertinent only within the authors' geographic locale, or did it have broader (even global) relevance?
- Were many other laboratories pursuing related research when the reported work was done? If so, why?
- For experimental research, what funding sources met the costs of the research?
- Was the selection of the research topic influenced by the source of research funding?
- On what prior observations was the research based? What was and was not known at the time?
- How important was the research question posed by the researcher?
# Body
Follow the structure of the journal article. Evaluate each section of the article – Introduction, Methods, Results, Discussion – highlighting the strengths and weaknesses of each section. Use the answers to the questions in Evaluating the Text to develop this section.
# Evaluating the Text
After you have read the article and answered the questions in the previous section, you should have a good understanding of the research undertaken. You can now begin to evaluate the author's research. Making judgments about someone else's work is often the most difficult part of writing the review. Many students feel that, because they are new to a discipline, they do not have enough knowledge to make judgments of other people's work.
# INTRODUCTION
- Read the statement of purpose in the introduction. What was the objective of the study?
- Consider the title. Does it precisely state the subject of the paper?
- Read the statement of purpose in the abstract. Does it match the one in the introduction?
- Check the sequence of statements in the introduction. Does all the information lead coherently to the purpose of the study?
# METHODS
- Review methods in relation to the objective(s) of the study. Are they valid for studying the problem?
- Check the methods for essential information. Could the study be duplicated from the methods and information given?
- Check the methods for flaws. Is the sample selection adequate? Is the experimental design sound?
- Check the sequence of statements in the methods. Does all the information belong there? Is the sequence of methods clear and pertinent?
# RESULTS
- Examine carefully the data as presented in the tables and diagrams. Does the title or legend accurately describe the content? Are column headings and labels accurate? Are the data organized for ready comparison and interpretation (a table should be self-explanatory, with a title that accurately and concisely describes content and column headings that accurately describe information in the cells)?
- Review the results as presented in the text while referring to the data in the tables and diagrams. Does the text complement, and not simple repeat, data? Are there discrepancies between the results in the text and those in the tables? Check all calculations and presentation of data.
- Review the results in light of the stated objectives. Does the study reveal what the researcher intended?
# DISCUSSION
- Check the interpretation against the results. Does the discussion merely repeat the results? Does the interpretation arise logically from the data or is it too far-fetched? Have the faults/flaws/shortcomings of the research been addressed?
- Is the interpretation supported by other research cited in the study?
- Does the study consider key studies in the field? Are there other research possibilities/directions suggested?
# OVERVIEW
- Reread the abstract. Does it accurately summarize the article?
- Check the structure of the article (first headings and then paragraphing). Is all the material organized under the appropriate headings? Are sections divided logically into subsections or paragraphs?
- Are stylistic concerns, logic, clarity and economy of expression addressed?
# General Checks
# Presentation quality
- Not just grammar, English and nice figures
- Articulation and presentation of arguments
# Technical quality
- Is the methodology sound?
- Are the conclusions credible?
TIP
Do not forget to mention the good
# Critical Review of a Paper Marking Rubric
- What is the paper about? ___/5%
Look at the abstract
- What are the author(s) trying to say/present? In other words what is the main message: A new theory? New data on a product? New manufacturing process? New methodology of measurement? ___/20%
Look at the contribution section, if any.
What is the evidence? (Measurements in the lab? Pilot plant? New formulae? Testing of a new process design? New instrumental techniques? ___/15%
Do you trust the evidence? Is there good reproducibility? Is the data reliable? Is the methodology sound? ___/10%
Have the author (s) succeeded in conveying their message? Are you convinced? ___/5%
Are the arguments clear and targeted? Is the proof complete? ___/5%
What are the scientific implications? Even, if it is not completely original does the paper still contribute new aspects or views in a field already well covered such as: Modifications of views; More in depth measurements of a physical, chemical, textural or sensory property; Modified methodology of measurement; Modified process or equipment. Can you use the methodology in this paper for your own work? ___/15%
What are the business implications, if any? ___/5%
Is the critical review well written? Clear presentation; Good standard of English; Good style; Concise
# Paper A: U-Net - Convolutional Networks for Biomedical Image Segmentation
# Introduction
This is a paper by Ronneberger et al., 2015, published by Springer International Publishing, Switzerland. The research was conducted with the aim of providing a network and training strategy that would use smaller datasets more efficiently to successfully train a deep network. The authors present the U-net architecture that is a convolutional neural network used for image segmentation that was first used for biomedical images. This deep network used data augmentation to amplify annotated available datasets, these augmentations were achieved using techniques such as rotation and deformations.
# Evidence
The architecture was tested in the International Symposium on Biomedical Imaging (ISBI) in 2015 and achieved a result almost 10% better than the second best. The network was also tested in three different segmentation tasks, although the segmentation map of one of them were kept secret at the time, so the reproducibility of the experiment was compromised. Nevertheless, more experiments in different areas rather than just medical would have been useful, since the authors stated that they were sure the network can be used for many more different applications. The results and the credibility of the ISBI and datasets used were enough to satisfy the expected result. Since the size of the datasets were small and the performance results were satisfactory, the authors we able to accomplish the objective, for biomedical segmentation tasks.
# Scientific Implications
The paper was relevant because of the use of smaller datasets to train the model; major models with high performance often use huge dataset to achieve the level of performance required on the contests, although in real applications, to produce a dataset is a very daunting task and sometimes even not possible. Recent studies have reported that the U-shaped networks made a profound impact on medical image segmentation [1].
# Business implications
For medical image analysis U-shaped networks have been very successful and used in many medical applications, the challenge of limited numbers of images in medical applications is one of the biggest of this field [1].
For weld identification U-Net is a very good choice for the size of the dataset necessary for the training; also the main problem with a weld path follower (the aim of the proposed project) is to identify if the image is a weld or not, therefore, image segmentation applies. For this kind of application, the most successful current commercial solutions make use of laser systems to track the weld bead, this makes the system more expensive than a machine vision based system, also the opportunity for improvement is greater on machine vision systems, with applications such as defect identification for instance.
# Paper B: Rigorous Tracking of Weld Beads for the Autonomous Inspection with a Climbing Robot
# Introduction
This paper was created by Terres et al., 2019, published in two different symposiums and in a workshop, respectively, the 2019 Latin American Robotics Symposium (LARS), 2019 Brazilian Symposium on Robotics (SBR) and 2019 Workshop on Robotics in Education (WRE). The main subjects of the paper are autonomous climbing robots, weld bead recognition, and fuzzy logic control design. The aim of this paper is to present the development of an autonomous climbing robot for oil tank inspection; the robot main features are to identify and track the weld bead using a laser profiler, and control the movement with a fuzzy logic controller that keeps the prototype on track.
# Evidence
The authors have fabricated a steel plate to test the robot capability, although, the plate used to simulate a spherical storage tank was not of representative significance since there are many different welding process that would generate a different weld bead profile. The robot successfully followed the weld bead on the test sample, this test is reproducible as long as the test piece is available. The prototype was not tested in remote access or great heights, which are the places where it would be much used in a real situation; the scene used to validate the work can be easily inspected manually without the need of an autonomous robot.
# Scientific Implications
The paper itself is not of great impact in the scientific community, and there are no huge breakthroughs with the techniques used; nevertheless, there are not many works developed involving climbing inspection robots that follows weld beads, which makes the work rendered very important for the development of such area. The fuzzy logic control system can be used for the proposed work with the variation of the weld bead input, since the weld bead localization will be made by a computer vision system using deep neural networks.
# Business implications
The concept can be used on commercial applications, although the bulkiness of the robot and the tall profile doesn’t allow for market readiness; the manoeuvrability of the robot could also be improved as well as the accuracy of the controller. The laser application is already used in the market as a successful implementation. There are not autonomous robots for welding inspection currently available for the NDT (Non-Destructive Testing) market.
# Paper C: A structured light vision sensor for on-line weld bead measurement and weld quality inspection
# What is the paper about?
# What are the author(s) trying to say/present?
# Evidence
# Do you trust the evidence? Is there good reproducibility? Is the data reliable? Is the methodology sound?
# Have the author (s) succeeded in conveying their message? Are you convinced?
# Are the arguments clear and targeted? Is the proof complete?
# The good
The paper is of relevance because of the use of smaller datasets to train the model, major models with high performance often use huge dataset to achieve the level of performance required on the contests, although in real applications, to produce a dataset is a very daunting task and sometimes even not possible.
# Scientific Implications
# Can you use the methodology in this paper for your application?
For weld identification the U-Net is a very good choice because of the size of the dataset and because the problem with a path follower which detects weld is to know if what it is seen is a weld or not, therefore, semantic segmentation applies.
# Business implications
For this kind of application the most successful current commercial solutions make use of laser systems to track the weld bead, this makes the system more expensive than a machine vision based system, also the opportunity for improvement is greater on machine vision systems, like defects identification for instance.