01_training_script

Bear Classifier

Original bear_classifier.ipynb file of the Bear Classifier project.

Create dataset using google images and classify between teddy bear, grizzly bear and black bear using Pytorch and fastai.

Run in Jupyter Notebook file called bear_classifier.ipynb if you have computer with modern GPU or upload and run it in Google Colab if you do not have modern GPU.

Replace the bear dataset with your own image dataset. Retrain a resnet34 image classification model. Such retraining is called transfer learning.

Save the resulting model pkl file to google drive, save the download link.

List of URLs to download the respective class images:

• blackbear.txt
• grizzlybear.txt
• teddybear.txt

To start transfer learning on the local computer (you need modern GPU for it):

# Install additional packages
pip install jupyter
# Run it
jupyter notebook
# Or run it via Google Colab

Shared link with the bear_classifier.ipynb script.