Inspiration
We love cats. The internet loves cats. But sometimes we have so many pictures of our fluffy friends that we don't know which one to share! So how do we choose?
What it does
KatClick allows you to upload your own cat picture, which it will then use to calculate a projected engagement score for your picture. Just upload the image, enter your follower count, and the backend will work its magic to give you an estimated number of likes for the photo and a rating (Bad, Below Average, Average, Above Average, Good) for how the photo will compare to others.
How I built it
The frontend was made with React. The backend was created by training a convolutional neural network for regression using Keras and Tensorflow with #cat images we scraped from Instagram. Scraped images were also passed through Google Cloud's Vision API to filter all pictures that weren't actually cats.
Accomplishments that I'm proud of
Actually scraping successful data despite Instagram's best attempts to thwart us. Actually creating a convolutional neural network that was able to give somewhat successful regression values. Actually making a fun, creative frontend to top it off.
What I learned
That the hardest part of a project is merging all the different components together. We thought we were quite on-time after having finished the frontend and backend, but a whole new host of problems arose in trying to merge them together.
We all learned lots of new skills, frameworks, and technologies during this wonderful experience. Our frontend team with React, backend with Python scripts using Google Cloud, TensorFlow, and BeautifulSoup for the scraper.
What's next for KatClick
KatClick is more of a proof-of-concept than anything else. There is a massive amount of interest in maximizing social media attention, a task that can be optimized using machine learning. Cat pictures are cute, but the same technology could be used for, say, businesses to gauge which social media campaign to choose or activists trying to maximize social impact for their causes..
Built With
- css
- google-cloud
- html
- keras
- python
- react
- tensorflow
Log in or sign up for Devpost to join the conversation.