Problem Statement 1: Neural Style Transfer
Imagine a world where you could transform any ordinary photograph into a masterpiece by
applying the artistic style of renowned painters like Van Gogh, Picasso, or Monet. Your task
is to develop a neural style transfer model that leverages state-of-the-art deep learning
techniques to merge the content of one image with the style of another, creating stunning
new visuals that blend the two seamlessly.
This model should be capable of extracting the stylistic features from a given artwork and
applying them to a different image while preserving the original content's structure and
details. The aim is to achieve a balance between the content and style, resulting in images
that are visually appealing and artistically coherent.
You are encouraged to explore techniques such as convolutional neural networks (CNNs),
optimization algorithms, and perceptual loss functions to accomplish this. However, the
use of external APIs is not permitted.
You have the freedom to choose how to implement and showcase your model. Options
include creating a web interface where users can upload images and select styles to apply
or integrate the model into existing software tools for artistic creation.
Here are some references to help you get started:
https://arxiv.org/pdf/1508.06576.pdf https://arxiv.org/pdf/1603.08155.pdf
https://arxiv.org/pdf/1703.06868.pdf https://arxiv.org/pdf/1705.06830.pdf
https://arxiv.org/pdf/1804.03547.pdf https://arxiv.org/pdf/1912.07921.pdf
Your goal is to develop a model that not only produces visually striking
images but also demonstrates a deep understanding of both the content
and the artistic style, enabling users to create unique and personalized
artwork.
Submission Form PS1: https://forms.gle/FGGrojBHPZgJmYWT6
WhatsApp Group - https://chat.whatsapp.com/GVRpafvCWykCv6dBIF3WJM
Problem Statement 2: Text to Image Generation
Imagine a world where you could describe an image in natural language, and an AI could
generate that image for you with stunning accuracy and detail. Your task is to develop a
text-to-image generative model that leverages state-of-the-art natural language processing
techniques to create realistic images based on textual descriptions.
This model should be capable of understanding and translating textual descriptions into
high-fidelity images, capturing intricate details and nuances while maintaining coherence
with the input description. It should be adaptable to various domains, from everyday
objects to complex scenes and landscapes.
You are encouraged to explore techniques such as attention mechanisms, transformer
architectures, and adversarial training to achieve this. However, the use of external APIs is
not permitted.
You have the freedom to choose how to implement and showcase your model. Options
include creating a web interface where users can input text descriptions and receive
corresponding images or integrating the model into existing software tools.
Here are some references to help you get started:
https://arxiv.org/pdf/1511.02793.pdf https://arxiv.org/pdf/1609.04802.pdf
https://arxiv.org/pdf/1802.06454.pdf https://arxiv.org/pdf/1812.04948.pdf
https://arxiv.org/pdf/2104.04843.pdf https://papers.nips.cc/paper/6125-
improved-techniques-for-training-gans.pdf
https://distill.pub/2018/differentiable-parameterizations/
https://www.coursera.org/specializations/generative-adversarial-
networks-gans
Your goal is to develop a model that not only generates visually appealing images but also
demonstrates a deep understanding of the textual descriptions, opening new possibilities
for creative expression and content creation.
Submission Form PS2: https://forms.gle/EdrbafxvuvvX8ZtN9
WhatsApp Group - https://chat.whatsapp.com/CHMT3BKV8SnHBni08Hxoq4
Problem Statement 3: Pdf Answering AI
Have you ever wondered if your pdf could talk to you and respond to your questions and
queries? With this project you can create an AI that responds to your questions about the
pdf that is being viewed at the moment.
Your task at hand is to create a model that leverages natural language processing
techniques to provide quick and accurate responses from the context of the pdf that is
being viewed. You are free to use any methods within the constrictions of the project like
word2vec, glove embedding technique etc.., however, use of APIs is not allowed.
There are multiple ways to use your model and display it's working. You are free to use any
method. Following is some of the methods recommended to use:
- Create a google extension (brownie points)
- Using a simple upload pdf and question interface (recommended)
Following are some references to help you get started:
- https://arxiv.org/pdf/1805.08092.pdf
- https://ieeexplore.ieee.org/abstract/document/9079274
- https://arxiv.org/pdf/1707.07328.pdf
- https://arxiv.org/pdf/1810.04805.pdf
- https://papers.nips.cc/paper/7181-attention-is-all-you-need.pdf
- https://nlp.seas.harvard.edu/2018/04/03/attention.html
- http://jalammar.github.io/illustrated-transformer/
- https://jalammar.github.io/visualizing-neural-machine-translation-mechanics-
ofseq2seq-models-with-attention/
- https://www.coursera.org/learn/nlp-sequence-models
- https://www.youtube.com/playlist?list=PLam9sigHPGwOBuH4_4fr-XvDbe5uneaf6
Submission Form PS3: https://forms.gle/wfxbKzpFvcYLqc8r7
WhatsApp Group - https://chat.whatsapp.com/IgRkDwpDFPX6ZgyChzHr4Y
*Submission deadline for all three PS is 20th June 2024, 11:59 pm.
Guidelines: -
To ensure the quality and originality of submissions, please adhere to the following
guidelines when submitting your machine learning project for verification:
1. Originality and Integrity
• No Copied Projects: Submissions must be original. Any form of plagiarism will result
in immediate disqualification.
• Source Verification: If your project is found to be copied or significantly derived from
existing work without proper attribution, it will be discarded without review.
2. GitHub Repository
• Project Upload: Upload your entire project to a GitHub repository. Ensure that all
necessary files are included.
• Repository Structure: Organize your repository clearly. Include folders for code,
data, results, and documentation.
• README File: Include a comprehensive README file with the following:
- Project Overview: A brief description of your project.
- Installation Instructions: Steps to set up the environment and run the
project.
- Usage: Examples of how to use the code.
- Dependencies: List of all required libraries and dependencies.
3. Detailed Report
• Length and Format: Submit a detailed report of 7-10 pages in PDF format.
• Content Requirements:
- Introduction: Explain the problem statement and the objectives of your project.
- Approach: Describe your methodology and the steps you followed to solve the
problem.
- Failed Approaches: Document any approaches that did not work, along with
reasons for their failure.
- Results: Present the results of your project, including any relevant metrics,
graphs, or visualizations.
- Discussion: Analyze the results, discussing their significance and any insights
gained.
- Conclusion: Summarize your findings and suggest possible future
improvements.
- References: Include citations for any external resources or papers you referred
to.
4. Submission Process
• Deadline: Ensure your project is submitted by the specified deadline.
• Verification: All submissions will undergo a thorough review process to verify
originality and quality and will be verified by ArIES upon completion.
Thank you for your participation!
