Skip to content

3.Project Submission

Marsha Gómez edited this page Jan 4, 2021 · 1 revision

Project submission

Unless otherwise stated:

  • The project must be submitted by sharing the Drive folder ("Visualization" mode) with the following accounts
  • The project must be submitted few days (typically 3-5) before the date of the oral exam.
  • A part of the oral examination will be dedicated to the discussion of the project.

The shared folder must be organized as follows:

|__Surname[_Surname2[_Surname3]]
    |____ Scratch_CNN.ipynb # one or more notebook
    |____ Pretrained_CNN.ipynb # one or more notebook
    |____ BaselineCNN.ipynb # one or more notebook
    |____ Ensemble.ipynb # one or more notebook
    |____ ... (other .ipynb, if needed)
    |____ Submission.csv
    |____ Report.pdf
    |____ models
          |______ store your models checkpoints here

In other words, you must include

  • all the notebook files (ipynb) with tasks solutions, adequately described and commented. The cells output should reflect the results reported in the report.

  • Network checkpoint models, stored in a dedicated folder.

  • Report.pdf. The report must include:

    • A description of the state-of-the-art/related works, with particular focus on the works/implementations that inspired your solutions.
    • A detailed description of the proposed solutions, highlighting the motivation behind the performed choices, with commented figures and plots.
      • E.g.:
        • simple model -> accuracy curves obtained on a validation set allows to observe underfitting
        • more complex model -> accuracy curves obtained on a validation set allows to observe overfitting
        • application of these regularization techniques: ...
        • performed these hyperparameters search: ...
        • hyperparameters chosen according to ...
      • details about how you produce your contest submission
  • File Submission.csv with the predictions of your best classifier (or composite classifier) evaluated on the private_test_images for the problem of diagnosis classification (benign v malignant). The submission guidelines can be found in the Notebook Contest.ipynb. The file must be comply with the following format, i.e. one label per line (0 for benign, 1 for malignant):

Project Result

Clone this wiki locally