There is nothing more unique than having a poem created from the interpretation of an image

What it does

It takes an image, URL or from the computer, runs it through Clarifai's API to generate a list of tags associated with that picture, then it grabs a multitude of lines of poems from a database, sees which combination is the best through IBM Watson, and chooses the best one according to an emotional parameter.

How we built it

We divided the work into making sure we can grab the list of tags from an arbitrary image, while the other person was taking care of the poem generation from dummy data. When things got connected, we expanded and got to polish everything up, and spent the rest of the time optimising the poem generation, through choosing the best lines of poems and having them judged by IBM Watson's Tone Analyser.

Challenges I ran into

Python proved to be challenging for us since we weren't experienced enough in it. We sometimes spent more time doing trivial stuff that should be accomplished immediately in a comfortable language. Nevertheless, practice makes perfect, and we never got impeded enough that moving forward wasn't an option anymore.

Accomplishments that we are proud of

We successfully glued together a remarkable number of APIs and functions all in a python script designed to synthesize awe from analytical data. Parsing JSON proved to be easier than expected, APIs credentials and authentication went surprisingly well. We managed to simulate meaningful emotion in the form of a poem from a picture of basically anything.

What I learned

Knowing your tools goes a long way, as well as having a really organised working environment that can support even the most difficult research sessions.

What's next for Poem from your picture

Implementing a beautiful web interface for easier access and spreading the word to anyone feeling that their picture isn't worth a thousand words. It always will.

Share this project: