Team members: Sarah Innis, Anthony Nguyen, Annie Staker, Izzy Valdivia
Cascadia Classifier is available here: Cascadia Classifier
Web-based tool
- How accurately can we classify landmarks?
- Can we distinguish between similar-appearing landmarks (e.g. same area)?
- Can we identify landmarks through different environments and lighting conditions?
- Can we identify landmarks if given a low-quality image input?
- Can we identify landmarks that we have not trained our model upon?
- Can we provide further information about a landmark?
Create a multiclass image classification tool that can correctly identify landmarks (tentatively within Washington state) if given an image.
- Images of Landmarks across the world, provided by Google
- More than 4 million labeled landmark photos
- Wikimedia
- (specifically, the Wikimedia link for a given landmark)
- Determine location information about a landmark
- Provide additional information about a landmark
- User-taken Photos of Landmarks in Washington
- Use to validate model & demonstrate model accuracy and precision
- Note: unzip these files in walandmarks/data
- Data for WikiMedia Scraping
- From the Google Landmarks Dataset
- Google Drive Link
- Data For Training Our Model
In order to create a conda environment from the current environment.yml file, run the following line:
conda env create -f environment.yml
To activate the environment run:
conda activate landmarks_classification_env
After configuring an environment, run the Streamlit app using this command:
python -m streamlit run walandmarks/ui/Home.py
Our unit tests can be found at walandmarks/test.
To run all unit tests:
python -m unittest discover
To see code coverage:
coverage run -m unittest discover
This repo uses PyLint as its linter. To run pylint:
PYTHONPATH=. pylint walandmarks