Group members: Adam Alder, Kevin Vulcano, Brandon Kolesar
This project used two datasets, both from Kaggle.
- Aircraft images dataset: https://www.kaggle.com/datasets/seryouxblaster764/fgvc-aircraft
- Bird images dataset: https://www.kaggle.com/datasets/gpiosenka/100-bird-species
There are three Jupyter Notebooks:
-
preprocessing.ipynb- Contains the code to preprocess the images. -
final-project.ipynb- Contains the code to train the CNN models and report on their accuracy. -
visualizations.ipynb- Contains the code to visualize the CNN.
The CNN models are defined in the models/CNN_model.py and models/CNN_SE_model.py files.
Due to performing the same analysis three times, once on only bird images, once on only aircraft images, and once on a combination of bird and aircraft images, the code and results were split up into different Git branches.
-
birds-only- The code modified to only work on bird images and its results. -
aircrafts-only(notaircraft-only) - The code modified to only work on aircraft images and its results. -
combined-data- The code unmodified and its results. Works on both bird and aircraft images.