This is a final project for the MIT Beaverworks Summer Institute class CogWorks. I built a tag generator that takes an image as input and outputs a handful of relevant hashtags. The model passes the image through ResNet to generate an image embedding and then an encoder to generate an embedding in the Word2Vec space. Finally the words with the closest embeddings are chosen as tags.
Note that several massive files are missing due to Github's size limit. Please see the three Jupyter notebooks to view the different components of the project. The final product can be seen in the display notebook or in the jpg titled Sheep Capper Example. Thank You!