PROJECT INSTRUCTIONS:
---------------------------------------------------------------------------
Your task for this project will be to train a model that can predict previously unseen classes.
A few points:

- In total, there are 200 classes, however the images for 25 of these classes have been removed and
will be used by us for testing.
- You are however given the names of all 200 classes, as you will need to identify attributes for the 
unseen classes in order to be able to classify them at testing time.
- We recommend using a python package (eg wikipedia-api) in order to scrape the web for attributes about
the bird species
- Out of the 175 classes we give you, we also reccomend that you reserve a certain amount to test on
as 'unseen' classes
- We will be running each notebook manually after submission in order to ensure that none of the removed
classes have been used for training
- We also require that you submit to gradescope a .py file which contains a function gradescope_predictions().
This function should take in a list of images, and return a list of predictions in the range [1-200]. (Note,
this function will need to handle both seen and unseen classes).
- While you are provided with the attributes that are present in each image, please note that these should
only be used when training your model. At testing time, it is the models job to predict the presence of these
attributes.

---------------------------------------------------------------------------
Below is the readme from the caltech birds dataset
---------------------------------------------------------------------------
The files in the attributes folder contain attribute labels obtained from Mechanical
Turk workers on the Birds-200 dataset. The file labels.txt contains the actual
labels. Each line corresponds to one attribute label of the form: 
<image_id> <attribute_id> <is_present> <certainty_id> <worker_id>

The file 'images.txt' contains lines of the form: 
<image_id> <image_file_name>

The file 'images-dirs.txt' is the same as 'images.txt' but also includes the
directory names: 
<image_id> <full_image_path>

The file 'attributes.txt' contains lines of the form: 
<attribute_id> <attribute_name>

<is_present> is 0 or 1, and indicates whether or not the worker thought the
given attribute was present in the given image.

The file 'certainties.txt' contains lines of the form: 
<certainty_id> <certainty_name>

Each unique value of worker_id corresponds to a unique worker on Mechanical
Turk.