Monash University, Department of Civil Engineering
CIV4100 Autonomous Vehicle Systems
ASSIGNMENT 2 RUBRIC
Criteria Poor (F) Average (P) Good (C) Very Good (D) Highest Quality (HD)
Task 1.1: Dataset
processing and
analysis (10%)
● The code is not working, or data is read
incorrectly
● The data is not divided correctly
● Part of the visualization is displayed
incorrectly
● The visualization contains incorrect
information or misses key information
about the dataset
● The code logic and organization are
unreadable or hardly understandable
● There is/are bugs in the code
● None or only a few parts of the code could
be reused in other programs.
● The code is brute force and unnecessarily
long.
● Findings are lacking or poorly presented
and organized without insights.
● The code works, data is read
successfully, but there may be 1-2 code
issues remaining.
● The visualization is displayed correctly
and meets basic specifications.
● Some information presented in the
visualization is redundant
● The code is fairly easy to read, has
sufficient code comments and somewhat
follows coding convention; and there may
be issues with the code.
● Findings are discussed to some extent
but lack depth or insight.
● Presentation of results is reasonably
clear and organized.
● The code works well with all data being
read correctly.
● The visualization meets all specifications.
● The visualization is readable, its font size
is the same as the font used in the report
● There is no bug or coding issue.
● The code is well organized, easy to read
with good code comments, and follows
coding conventions.
● The code could be reused as a whole
and each routine could be reused.
● Findings are adequately discussed with
insights.
● Presentation of results is well-structured
and organized.
● Everything in Good (C) plus:
● The visualization is correct and
consistent in terms of colors, sizes and
format with others in the report
● The code is exceptionally well organized
and strictly follows coding conventions.
● The code is efficient without sacrificing
readability and understanding.
● Findings are well discussed with good
insights.
● Presentation of results is highly
organized and coherent.
● Everything in Very Good (D) plus:
● The figures are professional and
perfectly annotated
● The code is outstandingly organized, with
professionally following coding
conventions.
● Findings are discussed
extensively/critically with depth/novel
insights.
● Code is compacted, optimized and
extremely efficient.
Task 1.2: Deep
learning model
development
(10%)
● The data processing and visualization are
either incorrect or poorly presented.
● The deep learning model produces
incorrect results, or correct results but does
not display them correctly.
● Only one or two performance measures are
used.
● The code logic and organization are
unreadable or hardly understandable.
● None or only a few parts of the code could
be reused in other programs.
● The code is brute force and unnecessarily
long.
● Findings are lacking or poorly presented
and organized without insights.
● The data processing and visualization
meet the majority of the specifications.
● The deep learning model works and
produces correct results and displays
them.
● Performance measures including means
of accuracy, precision, F1 are obtained
correctly.
● The code is fairly easy to read, has
sufficient code comments and follows
coding conventions.
● Findings/results are presented/discussed
to some extent but lack depth or insight.
● Presentation of results is reasonably
clear and organized.
● The data processing and visualization
meets all specifications.
● The deep learning model works well and
meets all of the specifications.
● The code is well organized, easy to read
with good code comments, and follows
coding conventions.
● The code could be reused as a whole
and each routine could be reused.
● There is no bug or coding issue.
● Findings are adequately discussed with
insights.
● Presentation of results is well-structured
and organized.
● Everything in Good (C) plus:
● The code is efficient without sacrificing
readability and understanding.
● The code is exceptionally well organized
and strictly follows coding conventions.
● The code is efficient without sacrificing
readability and understanding.
● Findings are well discussed with good
insights.
● Presentation of results is highly
organized and coherent.
● Everything in Very Good (D) plus:
● The code is outstandingly organized, with
professionally following coding
conventions.
● Code is compacted, optimized and
extremely efficient.
● Findings are discussed
extensively/critically with depth/novel
insights.
● Code is compacted, optimized and
extremely efficient.
Task 1.3: Model
improvement
(20%)
● The model produces incorrect results, or
correct results but does not display them
correctly.
● Code is not working or no noticeable
improvement in comparison with the model
in Task 1.2.
● The code logic and organization are
unreadable or hardly understandable.
● None or only a few parts of the code could
be reused in other programs.
● Methods/Findings are lacking or poorly
presented and organized without any
insight.
● The model works and produces the
correct results and displays them
correctly. It also meets most of the other
specifications.
● Better performance in comparison with
the model in Task 1.2.
● Rank and score in Kaggle have been
obtained correctly.
● The code is fairly easy to read, has
sufficient code comments and follows
coding conventions.
● Methods/Findings are
presented/discussed to some extent but
lack depth or insight.
● Presentation of results is reasonably
clear and organized.
● The model accuracy is ranked in the top
50% or higher in Kaggle.
● The model works well and meets all
specifications.
● There is no bug or coding issue.
● The code is well organized, easy to read
with good code comments, and follows
coding conventions.
● The code could be reused as a whole
and each routine could be reused.
● Methods/Findings are adequately
discussed with insights.
● Presentation of results is well-structured
and organized.

● Everything in Good (C) plus:
● The model accuracy is ranked in the top
20% or higher in Kaggle.
● The code is exceptionally well organized
and strictly follows coding conventions.
● The code is efficient without sacrificing
readability and understanding.
● Methods/Findings are well discussed
with good insights.
● Presentation of results is highly
organized and coherent.
● Everything in Very Good (D) plus:
● The model accuracy is ranked in the top
10% or higher in Kaggle.
● The code is outstandingly organized, with
professionally following coding
conventions.
● Code is compacted, optimized and
extremely efficient.
● Methods/Findings are discussed
extensively/critically with depth/novel
insights.
● Code is compacted, optimized and
extremely efficient.

Task 2.1: Testing
and validation
(15%)
● Some functions to generate test cases and
process data are incorrect or produce
wrong results.
● Some test cases are incorrect, or missing
test cases
● The testing is done incorrectly, or returns
incorrect results, or correct results but does
not display them correctly.
● Visualization is incorrect or buggy.
● The code logic and organization are
unreadable or hardly understandable.
● The code is brute force and unnecessarily
long; or the code is incomplete.
● Missing number of test cases.
● The code is brute force and unnecessarily
long.
● All functions to generate test cases and
process data are correct.
● Visualization is correct.
● There may be few issues or bugs with
the test cases and visualization.
● Both metamorphic testing and traditional
testing are implemented correctly.
● Metamorphic relations and test oracles
alongside their test cases are defined
correctly.
● Correct number of test cases are used,
and their results are displayed correctly
after the testing.
● The code and result in the report are
fairly easy to read, has sufficient
comments and code follows coding
conventions.
● Everything in Average (P) plus:
● The visualization is readable, its font size
is the same as the font used in the report
● Test cases are clearly and correctly
organized in the code and report with
correct information
● The code and result in the report are well
organized, easy to read with good
comments.
● There is no bug or coding issue.
● The code could be reused as a whole
and each routine could be reused.

● Everything in Good (C) plus:
● Test cases and their results are well
organized and follows conventions.
● The code and result in the report are
exceptionally well organized with good
insights, and code strictly follows coding
conventions.
● The code is efficient without sacrificing
readability and understanding.
● Everything in Very Good (D) plus:
● Test cases and their results are
exceptionally well organized and strictly
follows conventions.
● The code and result in the report are
outstandingly organized with new and
critical insights.
● Code is compacted, optimized and
extremely efficient.
Task 2.2:
Adversarial
attack (10%)
● Adversarial attack is implemented
incorrectly or unsuccessfully
● Missing visualization or demonstration of
successful attacks
● The code is incomplete or contains critical
errors.
● Implementation lacks clarity, organization,
or is excessively lengthy.
● Code logic is confusing or hard to follow.
● Code does not adhere to coding
conventions or lacks comments.
● Evaluation of the adversarial attack is
incomplete, incorrect, or lacks depth.
● Lack of clarity or organization in the results
and evaluations.
● Adversarial attack is implemented
correctly and successfully but may
contain some human errors.
● Code is somewhat clear and organized,
with some comments.
● There may be minor issues or bugs in
the implementation.
● Adheres to coding conventions to a
satisfactory level.
● Evaluation of the adversarial attack is
accurate but lacks depth or insight.
● Presentation of results is reasonably
clear and organized.

● Adversarial attack implementation is
well-structured and clear.
● The success of attack is limited by
low benchmark in the results
● Code is easy to understand and
well-commented.
● Minor issues or bugs are
addressed, and coding conventions
are followed.
● Comprehensive analysis of the
attack's effectiveness is provided.
● Presentation of results is well-
structured and organized.
● Some analysis of the attack's
effectiveness is provided.
● Everything in Good (C) plus:
● The success of attack is
demonstrated by highly successful
and sufficient results
● Code is well-commented and adheres
strictly to coding conventions.
● No issues or bugs are present, and the
code is highly reusable.
● Evaluation of the adversarial attack is
thorough and insightful.
● Presentation of results is highly
organized and coherent.
● Everything in Very Good (D) plus:
● The attack is successful with outstanding
results.
● The code is outstandingly organized,
with professionally following coding
conventions.
● Methods/Findings are discussed with
exceptional clarity, depth, and insight.
● Presentation of results is outstandingly
organized and coherent.
Task 2.3:
Defense against
adversarial
attack (15%)
● Implementation of defense methods is
incomplete or contains critical errors.
● Missing visualization or demonstration
of successful defense
● Chosen defense methods are
ineffective or inappropriate for the
task.
● Lack of clarity, organization, or
excessive length in implementation.
● Code logic is confusing or hard to
follow.
● Code does not adhere to coding
conventions or lacks comments.
● Evaluation of defense methods'
performance is incomplete, incorrect,
or lacks depth.
● Findings are not discussed or are
poorly articulated.
● Lack of clarity or organization in
presenting the evaluation results.
● Only 1 defense method is
implemented
● Defense method is implemented
correctly but may contain some
errors.
● Chosen defense method is
somewhat effective but could be
improved.
● Code is somewhat clear and
organized, with some comments.
● There may be minor issues or bugs
in the implementation.
● Adheres to coding conventions to a
satisfactory level.
● Evaluation of defense method's
performance is accurate but lacks
depth or insight.
● Findings are discussed to some
extent but lack depth or insight.
● Presentation of results is
reasonably clear and organized.
● Two different defense methods are
implemented correctly but may
contain some errors or issues in the
method, implementation, or result
● Chosen defense methods are
somewhat effective but could be
improved.
● Code is somewhat clear and
organized, with some comments.
● There may be minor issues or bugs
in the implementation.
● Adheres to coding conventions to a
satisfactory level.
● Evaluation of defense methods'
performance is accurate but lacks
depth or insight.
● Findings are discussed to some
extent but lack depth or insight.
● Presentation of results is
reasonably clear and organized.
● Two different defense methods are
implemented which is well-
structured and clear.
● Chosen defense methods are highly
effective in improving robustness
against adversarial attacks.
● Code is well-commented and
adheres strictly to coding
conventions.
● No issues or bugs are present, and the
code is highly reusable.
● Comprehensive analysis of defense
methods' performance is provided.
● Presentation of results is highly
organized and coherent.
● Detailed analysis of defense
methods' effectiveness is provided.
● Everything in Very Good (D) plus:
● Two different defense methods are
implemented which is exceptionally
clear, organized, and concise.
● Chosen defense methods are extremely
effective in improving robustness against
adversarial attacks.
● The defense can be applied to a wide
range of adversarial attacks
● Code is exceptionally well-commented,
organized, and adheres strictly to coding
conventions.
● No issues or bugs are present, and the
code is highly reusable.
● Evaluation of defense methods'
performance is exceptionally thorough,
insightful, and comprehensive.
Task 3:
Reporting &
recording (20%)
● Missing contents in the code, the
report or the video.
● There are no insightful discussions on
the results; and the content has a
fragmented and inconsistent structure
that doesn’t necessarily follow logic.
● Many subsections are very difficult to
understand; no interconnection exists
between tasks and their subsections,
making it difficult for the readers to
piece together the narrative.
● The content is not concise (a rambling
report within or over the page limit), formats
(including font size, heading, etc.) are
inconsistent
● Poor video and report presentation, not
professional. The recording is excessively
long.
● There may be few insightful discussions,
but they are weak. It also should have
basic logical structure but there remains
inconsistency in the content.
● The contents in some tasks are only
supported by a few graphics and
diagrams. Some tasks are difficult to
understand; little to no interconnection
exists between tasks and their
subsections, making it difficult for the
readers to piece together the narrative.
● Some contents are not concise and do
not follow the format and specification.
● Video and/or report presentation can be
improved.
● Everything in Average (P) plus:
● The report and recording are well
organized and there is no inconsistency.
● Few interconnections exist between
tasks and their subsections, but it needs
improvement (increases difficulty for the
readers to piece together the narrative).
● Content is reasonably concise and clear
(within the page limit), the format is
consistent throughout the report.
● Some revisions of grammar or spelling,
diction, syntax are still needed
● Good, concise video and report
presentation. The length of the recoding
is within the time limit.
● Everything in Good (C) plus:
● New insights are brought into the report.
● The report and video are exceptionally
well organized with high quality in format.
● Content is fairly concise and clear (within
the page limit), and consistent format.
● Good writing, with no need for revision of
grammar or spelling, diction, syntax.
● Excellent video and report presentation.
● Everything in Very Good (D) plus:
● Excellent writing and narrative, with no
need for revision of grammar or spelling,
diction, syntax
● Outstanding video and report
presentation.