7 PM – AI / ML – Shawn

Gamas Sub

Today We Did

1. We transferred Cat vs Dog project from Google Colab to Kaggle.
2. We reviewed the Cat vs Dog codes because it has been a while.
3. We enabled GPU accelerator for both Jayden and Theo. Initially their GPU accelerator were not working because their phone numbers were not verified.
4. We learned how to export the PKL file in Kaggle. This is because in Kaggle, there is no google drive integration. We need to do learn.export(‘/kaggle/working/my_model.pkl’) and then you can download the my_model.pkl from the right column into your computer.
5. We learned how to test the trained model in Kaggle by first uploading the data to Kaggle and do the learn.predict(“<kaggle uploaded file path”) . This is because in Kaggle below code does not work anymore

   1. from google.colab import files
      uploaded_file = files.upload()

6. We learned how to add other people’s dataset into their own Kaggle project.
7. We learned how to check if other people’s dataset is a good data set for computer vision or not.
8. You can find my Cat Vs Dog Kaggle project in https://www.kaggle.com/code/ayclogic/tue-7pm-cat-vs-dog

Homework

1. Each student will have to do a final project similar to Cat vs Dog streamlit project. It is due on 02/06.
2. For next week, student has to find a computer data set in Kaggle for their final project and show it to Shawn.
3. Since most of other people computer vision data set in Kaggle is not based on first letter of the file name (like the cat vs dog), you probably have to use something else other than ImageDataLoaders.from_name_func you probably have to use https://docs.fast.ai/vision.data.html#imagedataloaders.from_folder or  https://docs.fast.ai/vision.data.html#imagedataloaders.from_path_func. Please look in the for these 2 ImageDataLoaders and see if you can understand them or play around with your codes.