代写CS 417编程、代做Python程序语言

项目预算：开发周期：发布时间：要求地区：

CS 417/517: Introduction to Human Computer Interaction -
Project 1 ( Fall 2024 )
1 Introduction
In this assignment, your task is to implement a Convolutional Neural Network (CNN) and evaluate
its performance in classifying handwritten digits. After completing this assignment, you are able to
understand:
• How Neural Network works? How to implement Neural Network?
• How to setup a Machine Learning experiment on public data?
• How regularization, dropout plays a role in machine learning implementation?
• How to ffne-tune a well-train model?
To get started with the exercise, you will need to download the supporting ffles and unzip its
contents to the directory you want to complete this assignment.
2 Dataset
The MNIST dataset consists of a training set of 60000 examples and a test set of 10000 examples.
All digits have been size-normalized and centered in a ffxed image of 28 × 28 size. In the original
dataset, each pixel in the image is represented by an integer between 0 and 255, where 0 is black,
255 is white and anything between represents a different shade of gray. In many research papers, the
offfcial training set of 60000 examples is divided into an actual training set of 50000 examples and a
validation set of 10000 examples.
3 Implementation
( Notice : You can use any library to ffnish this project. We recommend students to use Google
Colab, which is a cloud-based service that allows you to run Jupyter Notebooks for free. To start
1this, follow these steps. 1. Open your web browser and go to the Google Colab website by visiting
colab.research.google.com. 2. Sign up and Sign in. 3. After signing in, you can start a new notebook
by clicking on File - New notebook. )
3.1 Tasks
Code Task [70 Points]: Implement Convolution Neural networks (CNN) to train and test the
MNIST or FER-2013 dataset, and save the well-train model.
Code Task (1) Build your customized Convolution Neural Network (CNN)
• Deffne the architecture of a Convolution Neural Network (CNN) with more than 3 layers, that
takes these images as input and gives as output what the handwritten digits represent for this
image.
• Test your machine learning model on the testing set: After ffnishing the architecture of CNN
models, ffx your hyper-parameters(learning rate, lambda for penalty, number of layers, and
number of neurons per layer), and test your model’s performance on the testing set.
• Implement different optimizer (i.e., at least two). Compare the results in report and analyze the
potential reasons.
• Implement different regularization methods for the Neural Networks, such as Dropout, l1 or l2.
Compare the results in report and analyze the potential reasons.
Code Task (2) Fine-tune at least three different well-pretrained models (e.g., MobileNetV3,
Resnet50 ) to get a good performance. You need to choose the speciffc layers to retrained and write
it in the report.
Code Task (3): This code task is only for CS517. Recognize handwritten digits from a
recorded video using the pre-trained model and OpenCV libraries.
Notice: The students in CS417 will get 20 points bonus if they ffnish this part.
Load the video and read frames.
Load the pre-trained model.
While the input is available, read the next frame.
Process the frame. (options: resizing, cropping, blurring, converting to
grayscale, binarizing, normalizing and etc.)
Input the processed frame into the model.
Use a threshold to detect digits.
Put a contour around the digit and label the predicted value and probability.
Display the frame.
Release resources.
Hint: Here lists some of the functions you might use.
cv2.VideoCapture
cv2.resize
cv2.cvtColor
2cv2.threshold
cv2.putText
cv2.rectangle
cv2.imshow
cv2.waitKey
cv2.destroyAllWindows
Writing Report Task [30 Points]: Write a report to describe above implementation details and
corresponding results.
4 Deliverables
There are two deliverables: report and code.
1. Report (30 points) The report should be delivered as a separate pdf ffle, and it is recommended
for you to use the NIPS template to structure your report. You may include comments in the
Jupyter Notebook, however you will need to duplicate the results in the report. The report
should describe your results, experimental setup, details and comparison between the results
obtained from different setting of the algorithm and dataset.
2. Code (70 points)
The code for your implementation should be in Python only. The name of the Main ffle should
be main.ipynb. Please provide necessary comments in the code and show some essential steps
for your group work.
3