Inspiration
Wanted to solve a real world problem. Matched images(submitted by user) to graffiti dataset.
What it does
It trains a model using 30000 images (collected from data.sfgov.org) to find if an image is graffiti or not .It then takes user uploaded input image and finds if it is a graffiti or not.
How I built it
Extracted all the graffiti dataset of 300000 rows. Analysed it and removed separate links to the graffiti url files. Used the graffiti url's to train a model on clarifai's infrastructure to show if an image is graffiti or not. Used clarifiai's api to train the model. Successfully trained the model.
Challenges I ran into
Challenges during extracting the dataset as its a big dataset. Then analysing it . Thereafter finding inconsistent data in the dataset was crucial for the outputs.
Accomplishments that I'm proud of
Combined bigdata analytics with webapp development to solve an interesting problem.
What I learned
I learned about image matching using concept tagging.
What's next for GrafittiMatch
I want to build a search engine around the dataset to provide better insights to users
Log in or sign up for Devpost to join the conversation.