一站式論文代寫,英国、美国、澳洲留学生Essay代寫—FreePass代写

python代寫--COMP9444:Assignment2
時間:2021-08-02

COMP9444 Neural Networks and Deep Learning

Term 2, 2021

Project 2 - Simpsons Character Classification

Due: Friday 6 August, 23:59 pm Marks: 30% of final assessment

Introduction

For this assignment you will be writing a Pytorch program that learns to classify 14 different Simpsons Characters using the grey scale images we provide.  


Example images of 9 different Simpsons Characters from the dataset.

Getting Started

Copy the archive hw2.zip into your own filespace and unzip it. This should create an hw2 directory containing the main file hw2main.py, configuration file config.py, skeleton file student.py and data zip file data.zip. You will need to unzip the data file in the same directory to create a subdirectory named data. Your task is to complete the file student.py in such a way that it can be run in conjunction with hw2main.py by typing python3 hw2 main.py

You must NOT modify hw2main.py in any way. You should only submit student.py(If you wish, you can modify config.py in order to switch between CPU and GPU usage)  

The provided file nw2main.py handles the following:

loading the images from the data directory

splitting the data into training and validation sets (in the ratio specified bytrain val split)

data transformation: images are loaded and converted to tensors: this allows the network to work with the data; you can optionally modify and add your own transformation steps, and you can specify different transformations for the training and testing phase if you wish

loading the data using DataLoader() provided by pytorch with your specified batch size in student.py

You should aim to keep your code backend-agnostic in the sense that it can run on either a CPU or GPU. This is generally achieved using the .to(device) function. If you do not have access to a GPU, you should at least ensure that your code runs correctly on a CPU.

Please take some time to read through hw2main.py and understand what it does.

Constraints

We have tried to structure hw2main.py so as to allow as much flexibility as possible in the design of your student.py code. You are free to create additional variables, functions, classes, etc., so long as your code runs correctly with hw2main.py unmodified, and you are only using the approved packages (i.e. those available on the CSE machines). You must adhere to these constraints:  

1. your model must be defined in a class called network.

2. the savedModel.pth file you submit must be generated by the student.py file you submit

3. make sure your version of pytorch and torchvision produce a savedModel.pth with the correct format to run on the CSE machines (which use rorch1.8.1 and rorchvisiono.9.1)

4. your submission (including savedModel.pth) must be under 50MB and you cannot load any external assets in the network class

5. while you may train on a GPU, you must ensure your model is able to be evaluated on a CPU

You must ensure that we can load your code and test it. This will involve importing your student.py file, creating an instance of your network class, restoring the parameters from your savedModel.pth, loading our own test dataset, processing according to what you specified in your student.py file, and calculating accuracy and score

You may NOT download or load data other than what we have provided. If we find your submitted model has been trained on external data you will receive zero marks for the assignment.

Question

At the top of your code, in a block of comments, you must provide a brief answer (about 300-500 words) to this Question:  

Briefly describe how your program works, and explain any design and training decisions you made along the way.

You should try to cover the following points in your Answer:

a. choice of architecture, algorithms and enhancements (if any)

b. choice of loss function and optimiser

c. choice of image transformations

d. tuning of metaparameters

e. use of validation set, and any other steps taken to improve generalization and avoid overfitting

Groups

This assignment may be done individually, or in groups of two students. Groups are determined by an SMS field called hw2group. Every student has initially been assigned a unique hw2group which is "h" followed by their studentID number, e.g. h1234567. If you plan to complete the assignment individually, you don't need to do anything (but, if you do create a group with only you as a member, that's ok too). If you wish to form a group, go to the COMP9444 WebCMS Page and click on "Groups" in the left hand column, then click "Create".Enter your Group Name and select the Group Type "hw2". After creating a Group, click "Edit" search for the other member, and click "Add". WebCMS assigns a unique group ID to each group, in the form of "g" followed by six digits (e.g. go12345). We will periodically run a script to load these values into SMS. You must ensure there are no more than two members in your group, and no-one is a member of two different groups .

 Submission

You should submit your trained model and Python code by typing

give cs9444 hw2 student.py savedModel.pth

You must submit your trained model savedModel.pth as well as the Python code student.pyIn order to avoid technical problems, you are strongly advised to submit from the command line, and not via the give Web interface. You can submit as many times as you like - later submissions by either group member will overwrite previous submissions by either group member. You can check that your submission has been received by using the following command:

9444 classrun -check

The submission deadline is Friday 6 August, 23:59. 15% penalty will be applied to the (maximum) mark for every 24 hours late after the deadline

Additional information may be found in the FAQ and will be considered as part of the specification for the project. You should check this page regularly

When you submit, the system will check that your model can be successfully loaded, anc evaluate it on data randomly chosen from a third dataset (disjoint from data. zip and also disjoint from the holdout test set).

Marking Scheme

After submissions have closed, your code wil be run on a holdout test set (i.e. a set of images and labels that we do not make available to you, but which we will use to test your model).Marks will be allocated as follows

14 marks for algorithms, design choices and answer to the Question 

2 marks for coding style and comments

14 marks based on performance on the (unseen) test set

The performance mark will be based on the accuracy of your network performance on the testing dataset, which is disjoint from the validation dataset we use to test your model at submission time.  

General Advice:

Use the variable train val split to help you make design decisions aimed to avoid overfitting to the training data. At the very end, you may wish to re-train using the entire training set.

Try to be methodical in your development. Blindly modifying code, looking at the output then modifying again can cause you go around in circles. A better approach is to keep a record of what you have tried, and what outcome you observed. Decide on a hypothesis you want to test, run an experiment and record the result. Then move on to the next idea.

You should consider the submission test script to be the final arbiter with regard to whether a certain approach is valid. If you try something, and the submission test runs and you get a good accuracy then the approach is valid. If it causes errors then it is not valid.

Do Not leave this assignment to the last minute. Get started early, and submit early in order to ensure your code runs correctly. Marks from automated testing are final. You should aim to be uploading your final submission at least two hours before the deadline.It is likely that close to the deadline, the wait time on submission test results will increase.  

Common Questions:

Can I train on the full dataset if I find it? No. You should NOT attempt to reconstruct the test set by searching the Internet. We will retrain a random selection of submissions, as well as those achieving high accuracy. If your code attempts to search or load external assets, or we find a mismatch between your submittied code and saved model, you will receive zero marks.

My model is only slightly larger than 50MB, can you still accept it? No, the 50MB limit is part of the assignment specification and is quite generous. You should be able to get away with much less.

Can we assume you will call net.eval () on our model prior to testing? Yes.  



在線客服

售前咨詢
售后咨詢
微信號
Essay_Cheery
微信
专业essay代写|留学生论文,作业,网课,考试|代做功課服務-PROESSAY HKG 专业留学Essay|Assignment代写|毕业论文代写-rushmyessay,绝对靠谱负责 代写essay,代写assignment,「立减5%」网课代修-Australiaway 代写essay,代写assignment,代写PAPER,留学生论文代写网 毕业论文代写,代写paper,北美CS代写-编程代码,代写金融-第一代写网 作业代写:CS代写|代写论文|统计,数学,物理代写-天天论文网 提供高质量的essay代写,Paper代写,留学作业代写-天才代写 全优代写 - 北美Essay代写,Report代写,留学生论文代写作业代写 北美顶级代写|加拿大美国论文作业代写服务-最靠谱价格低-CoursePass 论文代写等留学生作业代做服务,北美网课代修领导者AssignmentBack