search

bgorman3 / ITSC-3990-AI-Art-Authentication

New Research Project for ITSC 3990 Using AI to Authenticate Art
1 stars 0 forks source link

Create Issues #7

Closed btdobbs closed 2 weeks ago

btdobbs commented 2 weeks ago

Create research tasks given my two comments on 10/31 in #6.

bgorman3 commented 2 weeks ago

  1. First Task: Return to the original setup. I initially trained a model using 300 Monet and 300 non-Monet artworks, but I haven’t yet trained one using the full set of 1,366 Monet images you provided.

  2. Second Task: Create a balanced dataset with 1,366 non-Monet artworks from the 50 different artists in my collection.

  3. Third Task: Test this model using a Monet dataset it hasn’t seen before, plus a set of non-Monet images, to evaluate its performance.

  4. Fourth Task: Evaluate the model using 136 AI-generated images to match the size of the Monet test set.

  5. Fifth Task: (if time allows): Transition to a multi-class classification setup to introduce a third AI art category. This setup will yield better classification results, as the model will have knowledge of AI and Monet features and be required to determine the category it most likely belongs to.

  6. Sixth Task (if time allows): Set up an expanded multi-class classifier with 1,366 artworks from the 50 artists, plus an AI art category. This larger dataset should help the model distinguish features more effectively, allowing it to understand better and identify Monet's artwork.

Let me know if you think anything I have here isn't a good approach or what you would change.

btdobbs commented 2 weeks ago

Sorry, I said tasks but meant issues! I converted all tasks to issues with relative comments so I can close this issue.