首页
网站开发
桌面应用
管理软件
微信开发
App开发
嵌入式软件
工具软件
数据采集与分析
其他
首页
>
> 详细
CSC345编程代写、代做Python语言程序
项目预算:
开发周期:
发布时间:
要求地区:
CSC345/M45 Big Data and Machine Learning
Coursework: Object Recognition
Policy
1. To be completed by students working individually.
2. Feedback: Individual feedback on the report is given via the rubric within Canvas.
3. Learning outcome: The tasks in this assignment are based on both your practical
work in the lab sessions and your understanding of the theories and methods. Thus,
through this coursework, you are expected to demonstrate both practical skills and
theoretical knowledge that you have learned through this module. You also learn to
formally present your understandings through technical writing. It is an opportunity
to apply analytical and critical thinking, as well as practical implementation.
4. Unfair practice: This work is to be attempted individually. You may get help from
your lecturer, academic tutor, and lab tutor, but you may not collaborate with your
peers. Copy and paste from the internet is not allowed. Using external code
without proper referencing is also considered as breaching academic integrity.
5. University Academic Integrity and Academic Misconduct Statement: By
submitting this coursework, electronically and/or hardcopy, you state that you fully
understand and are complying with the university's policy on Academic Integrity and
Academic Misconduct.
The policy can be found at https://www.swansea.ac.uk/academic-services/academicguide/assessment-issues/academic-integrity-academic-misconduct.
6. Submission deadline: Both the report and your implemented code in Python need to
be submitted electronically to Canvas by 11AM 14
th December.
1. Task
The amount of image data is growing exponentially, due in part to convenient and cheap camera
equipment. Teaching computers to recognise objects within a scene has tremendous application
prospects, with applications ranging from medical diagnostics to Snapchat filters. Object
recognition problems have been studied for years in machine learning and computer vision
fields; however, it is still a challenging and open problem for both academic and industry
researchers. The following task is hopefully your first small step on this interesting question
within machine learning.
You are provided with a small image dataset, where there are 100 different categories of objects,
each of which has 500 images for training and 100 images for testing. Each individual image
only contains one object. The task is to apply machine learning algorithms to classify the testing
images into object categories. Code to compute image features and visualize an image is
provided, you can use it to visualize the images and compute features to use in your machine
learning algorithms. You will then use a model to perform classification and report quantitative
results. You do not have to use all the provided code or methods discussed in the labs so far.
You may add additional steps to the process if you wish. You are encouraged to use the
implemented methodology from established Python packages taught in the labsheets (i.e.
sklearn, skimage, keras, scipy,…). You must present a scientific approach, where you make
suitable comparison between at least two methods.
2. Image Dataset – Subset of CIFAR-100
We provide the 100 object categories from the complete CIFAR-100 dataset. Each category
contains 500 training images and 100 testing images, which are stored in two 4D arrays. The
corresponding category labels are also provided. The objects are also grouped into 20 “superclasses”. The size of each image is fixed at 32x32x3, corresponding to height, width, and colour
channel, respectively. The training images will be used to train your model(s), and the testing
images will be used to evaluate your model(s). You can download the image dataset and
relevant code for visualization and feature extraction from the Canvas page.
There are six numpy files provided, as follows:
• trnImage, 32x32x3x50000 matrix, training images (RGB image)
• trnLabel_fine, 50000 vector, training labels (fine granularity)
• trnLabel_coarse, 50000 vector, training labels (coarse granularity)
• tstImage, 32x32x3x10000 matrix, testing images (RGB image)
• tstLabel_fine, 10000 vector, testing labels (fine granularity)
• tstLabel_coarse, 10000 vector, testing labels (coarse granularity)
The data is stored within a 4D matrix, and for many of you this will be the first time seeing a
high dimensionality tensor. Although this can seem intimidating, it is relatively
straightforward. The first dimension is the height of the image, the second dimension is the
width, the third dimension is the colour channels (RGB), and the fourth dimension is the
samples. Indexing into the matrix is like as with any other numeric array in Python, but now
we deal with the additional dimensions. So, in a 4D matrix ‘X’, to index all pixels in all
channels of the 5th image, we use the index notation X[:, :, :, 4]. So, in a generic form, if we
want to index into the i,j,k,lth element of X we use X[i, j, k, l].
Figure 1. Coarse Categories of CIFAR-100 Dataset
aquatic mammals
fish
flowers
food containers
fruit and vegetables
household electrical devices
household furniture
insects
large carnivores
large man-made outdoor things
large natural outdoor scenes
large omnivores and herbivores
medium-sized mammals
non-insect invertebrates
people
reptiles
small mammals
trees
vehicles 1
vehicles 2
3. Computing Features and Visualizing Images
A notebook, RunMe.ipynb, is provided to explain the concept of computing image features.
The notebook is provided to showcase how to use the skimage.feature.hog() function to obtain
features we wish to train our models on, how to visualize these features as an image, and how
to visualize a raw image from the 4D array. You do not need to use this if your experiments
do not require it! You should also consider the dimensionality of the problem and the features
being used to train your models, this may lead to some questions you might want to explore.
The function utilises the Histogram of Orientated Gradients method to represent image domain
features as a vector. You are NOT asked to understand how these features are extracted from
the images, but feel free to explore the algorithm, underlying code, and the respective Python
package APIs. You can simply treat the features as the same as the features you loaded from
Fisher Iris dataset in the Lab work. Note that the hog() method can return two outputs, the first
are the features, the second is an image representation of those features. Computing the second
output is costly and not needed, but RunMe.ipynb provides it for your information.
4. Learning Algorithms
You can find all relative learning algorithms in the lab sheets and lecture notes. You can use
the following algorithms (Python (and associated packages) built-in functions) to analyse the
data and carry out the classification task. Please note: if you feed certain algorithms with a
large chunk of data, it may take a long time to train. Not all methods are relevant to the task.
• Lab sheet 2:
o K-Means
o Gaussian Mixture Models
• Lab sheet 3:
o Linear Regression
o Principal Component Analysis
o Linear Discriminative Analysis
• Lab sheet 4:
o Support Vector Machine
o Neural Networks
o Convolutional Neural Networks
5. Benchmark and Discussion
Your proposed method should be trained on the training set alone, and then evaluated on the
testing set. To evaluate: you should count, for each category, the percentage of correct
recognition (i.e., classification), and report the confusion matrix. Note that the confusion matrix
can be large, and so you may need to think of ways to present appropriately; you can place it
in your appendices if you wish, or show a particularly interesting sub-region.
The benchmark to compare your methods with is 39.43%, averaged across all 20 super
categories, and 24.49% for the finer granularity categories. Note: this is a reference, not a
target. You will not lose marks for being slightly under this target, but you should be aware of
certain indicative results (very low or very high) that show your method/implementation may
not be correct. Your report will contain a section in which you discuss your results.
6. Assessment
You are required to write a 3-page conference/publication style report to summarize your
proposed method and the results. Your report should contain the following sections:
1. Introduction. Overview of the problem, proposed solution, and experimental results.
2. Method. Present your proposed method in detail. This should cover how the features
are extracted, any feature processing you use (e.g., clustering and histogram generation,
dimensionality reduction), which classifier(s) is/are used, and how they are trained and
tested. This section may contain multiple sub-sections.
3. Results. Present your experimental results in this section. Explain the evaluation
metric(s) you use and present the quantitative results (including the confusion matrix).
4. Conclusion. Provide a summary for your method and the results. Provide your critical
analysis; including shortcomings of the methods and how they may be improved.
5. References. Include correctly formatted references where appropriate. References are
not included in the page limit.
6. Appendices. You may include appendix content if you wish for completeness,
however the content you want graded must be in the main body of the report.
Appendices are not included in the page limit.
Page Limit: The main body of the report should be no more than 3 pages. Font size should be
no smaller than 10, and the text area is approximately 9.5x6 inches. You may use images but
do so with care; do not use images to fill up the pages. You may use an additional cover sheet,
which has your name and student number.
Source Code: Your submission should be professionally implemented and must be formatted
as an ipynb notebook. You may produce your notebook either locally (Jupyter, VSCode etc.),
or you may utilize Google Colab to develop your notebook, however your submission must be
an ipynb notebook. Remember to carefully structure, comment, and markdown your
implementation for clarity.
7. Submission
You will be given the marking rubric in advance of the submission deadline. This assignment
is worth 20% of the total module credit.
Submit your work electronically to Canvas. Your report should be in PDF format only.
Your code must be in a .ipynb format. Both files should be named with your student number,
i.e. 123456.pdf and 123456.ipynb, where 123456 is your student number.
There are two submission areas on Canvas, one for the report and another for the .ipynb
notebook. You must upload both submissions to the correct area by the deadline.
The deadline for this coursework is 11AM 14
th December.
软件开发、广告设计客服
QQ:99515681
邮箱:99515681@qq.com
工作时间:8:00-23:00
微信:codinghelp
热点项目
更多
代写dts207tc、sql编程语言代做
2024-12-25
cs209a代做、java程序设计代写
2024-12-25
cs305程序代做、代写python程序...
2024-12-25
代写csc1001、代做python设计程...
2024-12-24
代写practice test preparatio...
2024-12-24
代写bre2031 – environmental...
2024-12-24
代写ece5550: applied kalman ...
2024-12-24
代做conmgnt 7049 – measurem...
2024-12-24
代写ece3700j introduction to...
2024-12-24
代做adad9311 designing the e...
2024-12-24
代做comp5618 - applied cyber...
2024-12-24
代做ece5550: applied kalman ...
2024-12-24
代做cp1402 assignment - netw...
2024-12-24
热点标签
mktg2509
csci 2600
38170
lng302
csse3010
phas3226
77938
arch1162
engn4536/engn6536
acx5903
comp151101
phl245
cse12
comp9312
stat3016/6016
phas0038
comp2140
6qqmb312
xjco3011
rest0005
ematm0051
5qqmn219
lubs5062m
eee8155
cege0100
eap033
artd1109
mat246
etc3430
ecmm462
mis102
inft6800
ddes9903
comp6521
comp9517
comp3331/9331
comp4337
comp6008
comp9414
bu.231.790.81
man00150m
csb352h
math1041
eengm4100
isys1002
08
6057cem
mktg3504
mthm036
mtrx1701
mth3241
eeee3086
cmp-7038b
cmp-7000a
ints4010
econ2151
infs5710
fins5516
fin3309
fins5510
gsoe9340
math2007
math2036
soee5010
mark3088
infs3605
elec9714
comp2271
ma214
comp2211
infs3604
600426
sit254
acct3091
bbt405
msin0116
com107/com113
mark5826
sit120
comp9021
eco2101
eeen40700
cs253
ece3114
ecmm447
chns3000
math377
itd102
comp9444
comp(2041|9044)
econ0060
econ7230
mgt001371
ecs-323
cs6250
mgdi60012
mdia2012
comm221001
comm5000
ma1008
engl642
econ241
com333
math367
mis201
nbs-7041x
meek16104
econ2003
comm1190
mbas902
comp-1027
dpst1091
comp7315
eppd1033
m06
ee3025
msci231
bb113/bbs1063
fc709
comp3425
comp9417
econ42915
cb9101
math1102e
chme0017
fc307
mkt60104
5522usst
litr1-uc6201.200
ee1102
cosc2803
math39512
omp9727
int2067/int5051
bsb151
mgt253
fc021
babs2202
mis2002s
phya21
18-213
cege0012
mdia1002
math38032
mech5125
07
cisc102
mgx3110
cs240
11175
fin3020s
eco3420
ictten622
comp9727
cpt111
de114102d
mgm320h5s
bafi1019
math21112
efim20036
mn-3503
fins5568
110.807
bcpm000028
info6030
bma0092
bcpm0054
math20212
ce335
cs365
cenv6141
ftec5580
math2010
ec3450
comm1170
ecmt1010
csci-ua.0480-003
econ12-200
ib3960
ectb60h3f
cs247—assignment
tk3163
ics3u
ib3j80
comp20008
comp9334
eppd1063
acct2343
cct109
isys1055/3412
math350-real
math2014
eec180
stat141b
econ2101
msinm014/msing014/msing014b
fit2004
comp643
bu1002
cm2030
联系我们
- QQ: 9951568
© 2021
www.rj363.com
软件定制开发网!