Project Report

 You should structure your project report like the research papers we have been reading and presenting in class. 

Project Report:

1. Abstract
2. Introduction
3. Related Work
4. Methods
5. Results and Analysis
6. Conclusion
7. Contribution
8. References
9. (OPTIONAL) Appendix which includes supplemental material

Please take the report seriously.  It is hard to give credit for all your hard work, if the work is not clearly described in the report.  In writing the report, make clear what your contributions were beyond what is available on the internet.  If you re-implemented a deep learning model from the literature, describe this. If you built your project from an existing implementation,  describe this. 

Please refer to this document (final_project_guidelines.pdf) for detailed guidelines on final report and its submisison. 

Please refer this document for Final Project Guidelines.


Due Dates:

Project Proposal: 19th Feb, 2025 (@ 2:10 pm, before class)
Team Presentations: 23rd April and 30th April, 2025
Final Report: 5th May, 2025
Video Presentation: 5th May, 2025

Resources

Software Resources (all one Google search away...):

• Python and Pytorch
• Jupyter Notebook
• Hugging Face
• ​GitHub
• Tensorflow
• Tensorboard​

Background on Deep Learning:

• Deep Learning, Goodfellow, Bengio, and Courville​

Background on Old-School Computer Vision:

• Computer Vision: A Modern Approach, Forsyth and Ponce
• Computer Vision: Algorithms and Applications, Richard Szeliski
• Receptive Fields, Binocular Interaction, and Functional Architecture in the Cat's Visual Cortex, Hubel and Wiesel, 1962
• Hubel and Weisel, Cat Experiments Video
• MIT Summer Vision Project, Papert, 1966
• ​Vision, Marr, 19
• Computer Vision, Ballard and Brown, 1982
• Robot Vision, Horn, 1985
• Pattern Classification, Duda, Hart, and Stork​
• Pegasos: Primal Estimated sub-Gradient Solver for SVM, Shalev-Shwartz et al.​​

Some Alternatives to GCP Cloud Coupons:
1. Google Colab (Free T4 GPU)

• Cost: Free
• Details: Provides limited GPU resources (T4) for sessions that can last up to 4 hours.
• Instructions:
  1. Go to colab.research.google.com.
  2. Sign in with your Google account.
  3. Create a new notebook and under Runtime → Change runtime type, select "T4 GPU" as the hardware accelerator.
• Considerations: The free version can disconnect or limit sessions unexpectedly, especially during peak usage. For more consistent performance and longer sessions, you can subscribe to Colab Pro or Pro+.

2. Microsoft Azure for Students ($100 Azure Credit)

• Cost: $100 in Azure credit, no credit card required for students who can verify their academic status.
• Details: Credit is valid for up to 12 months or until it’s used up.
• Instructions:
  1. Visit the Azure for Students page.
  2. Sign up with your university email address and verify your student status.
  3. Once activated, use Azure services (e.g. Virtual Machines, Azure Machine Learning) within your credit limit.
• Considerations: Exceeding the $100 credit or time limit means you may need to upgrade your account, which typically requires a credit card.

3. Amazon Web Services (SageMaker 2-Month Free Trial)

• Cost: 250 hours per month of t2.medium notebook usage for the first two months under a new AWS account.
• Details: Other AWS services outside of SageMaker’s free trial may incur charges.
• Instructions:
  1. Create a new AWS account at aws.amazon.com (credit card required for verification).
  2. Access Amazon SageMaker through the AWS Management Console.
  3. Launch a t2.medium notebook instance and track usage within free limits.
• Considerations: Carefully monitor usage to avoid accidental charges, as you will be billed for usage beyond free-tier limits or if you use additional AWS services.

4. Google Cloud Platform ($300 in Free Credit for New Customers)

• Cost: $300 credit for new GCP accounts.
• Details: Typically valid for 90 days, requiring a credit card for identity verification.
• Instructions:
  1. Go to cloud.google.com and create a new account.
  2. Provide credit card details for verification.
  3. Use the console to deploy services like Compute Engine, Vertex AI, or pre-configured deep learning VM images.
• Considerations: Keep track of your usage so you do not exceed the allotted $300 credit within the 90-day period, charges may apply thereafter.

5. IBM Cloud ($200 Cloud Credit)

• Cost: $200 credit for new IBM Cloud sign-ups.
• Details: Credit can be applied toward IBM services like Watson Machine Learning, and more.
• Instructions:
  1. Sign up for an IBM Cloud account at cloud.ibm.com.
  2. Complete any required verification steps (credit card may be requested).
  3. Begin using IBM Cloud services within the $200 credit limit.
• Considerations: The credit typically applies to your initial usage and expires after a set period. Any usage beyond the credit will be charged to your account. Also, you might have to wait for sometime for your account to be activated.

Note: Before committing to any platform, review the specific terms and policies for credits, free tiers, and expiration dates. Always monitor your usage to avoid unexpected charges, and set up usage alerts where possible.