Scribd
Upload
36 views4 pages

Project Title

This document outlines a project assignment for students to complete involving building models to classify or denoise images using deep learning techniques. Students can choose between three titles - using CNNs to classify CIFAR-10 images, using CNNs to denoise BSDS500 images, or using deep learning to denoise MNIST images. For each, tasks involve preprocessing data, building a model, training/evaluating it, and making predictions. Students must submit a Jupyter notebook with code and a report discussing their approach. The assignment will be graded based on various criteria such as data preprocessing and model performance.

Uploaded by

Iezma Madzin
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
36 views4 pages

Project Title

This document outlines a project assignment for students to complete involving building models to classify or denoise images using deep learning techniques. Students can choose between three titles - using CNNs to classify CIFAR-10 images, using CNNs to denoise BSDS500 images, or using deep learning to denoise MNIST images. For each, tasks involve preprocessing data, building a model, training/evaluating it, and making predictions. Students must submit a Jupyter notebook with code and a report discussing their approach. The assignment will be graded based on various criteria such as data preprocessing and model performance.

Uploaded by

Iezma Madzin
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

SKM3205 PROJECT QUESTIONS

DEADLINE: 19 June 2023 (Week 14)

This project will carry a total of 35% of your assessment. It is a group-work of 2


students per group. Choose only 1 of the following title which 3 groups will do the
same title.

Assignment Title 1: Image Classification using Convolutional Neural Networks


(CNNs)

Objective: The objective of this assignment is to build a model using CNNs to classify
images from the CIFAR-10 dataset. The CIFAR-10 dataset contains 60,000 32x32
color images in 10 classes, with 6,000 images per class. The classes are: airplane,
automobile, bird, cat, deer, dog, frog, horse, ship, truck.

Tasks:

1. Download the CIFAR-10 dataset and explore it to get an understanding of the


data.
2. Preprocess the data to get it ready for model training.
3. Build a CNN model using TensorFlow or Keras.
4. Train the model on the preprocessed data.
5. Evaluate the model on a test set to determine its accuracy.
6. Use the model to make predictions on new, unseen images.

Deliverables:

1. A Jupyter notebook with your code and comments explaining the code.
2. A report summarizing your findings and discussing your approach to building
the model. The report should include the following sections:
• Introduction
• Data Preprocessing
• Model Architecture
• Model Training and Evaluation
• Prediction on New Images
• Conclusion

Grading: Your assignment will be graded based on the following criteria:

• Data preprocessing (20 marks)


• Model architecture (20 marks)
• Model training and evaluation (40 marks)
• Prediction on new images (10 marks)
• Code readability and documentation (10 marks)
Note: You are allowed to use pre-trained models or transfer learning techniques to
improve your model performance, but you must explain your approach and provide
appropriate references.
Assignment Title 2: Image Denoising using Convolutional Neural Networks
(CNNs)

Objective: The objective of this assignment is to build a model using CNNs to


denoise images from the BSDS500 dataset. The BSDS500 dataset contains 500
natural images with various levels of noise added to them.

Tasks:

1. Download the BSDS500 dataset and explore it to get an understanding of the


data.
2. Preprocess the data to get it ready for model training.
3. Build a CNN model using TensorFlow or Keras.
4. Train the model on the preprocessed data with the noisy images as inputs and
the clean images as targets.
5. Evaluate the model on a test set to determine its denoising performance.
6. Use the model to denoise new, unseen images.

Deliverables:

1. A Jupyter notebook with your code and comments explaining the code.
2. A report summarizing your findings and discussing your approach to building
the model. The report should include the following sections:
• Introduction
• Data Preprocessing
• Model Architecture
• Model Training and Evaluation
• Image Denoising Results
• Conclusion

Grading: Your assignment will be graded based on the following criteria:

• Data preprocessing (20%)


• Model architecture (20%)
• Model training and evaluation (40%)
• Image denoising results (10%)
• Code readability and documentation (10%)

Note: You are allowed to use other denoising techniques in addition to CNNs, such
as wavelet transforms or non-local means filtering, but you must explain your
approach and provide appropriate references.
Assignment Title 3: Image Denoising using Deep Learning

Objective: The objective of this assignment is to build a model using deep learning to
denoise images from the MNIST dataset. The MNIST dataset contains 60,000 training
images and 10,000 test images of handwritten digits with various levels of noise
added to them.

Tasks:

1. Download the MNIST dataset and explore it to get an understanding of the


data.
2. Preprocess the data to get it ready for model training.
3. Build a deep learning model using TensorFlow or Keras.
4. Train the model on the preprocessed data with the noisy images as inputs and
the clean images as targets.
5. Evaluate the model on a test set to determine its denoising performance.
6. Use the model to denoise new, unseen images.

Deliverables:

1. A Jupyter notebook with your code and comments explaining the code.
2. A report summarizing your findings and discussing your approach to building
the model. The report should include the following sections:
• Introduction
• Data Preprocessing
• Model Architecture
• Model Training and Evaluation
• Image Denoising Results
• Conclusion

Grading: Your assignment will be graded based on the following criteria:

• Data preprocessing (20%)


• Model architecture (20%)
• Model training and evaluation (40%)
• Image denoising results (10%)
• Code readability and documentation (10%)

Note: You are allowed to use other deep learning techniques in addition to CNNs,
such as autoencoders or denoising autoencoders, but you must explain your
approach and provide appropriate references.

You might also like