UnpossiblyMachineLearning_Instagram

Unpossibly Machine Learning Competition

Welcome to the Instagram Machine Learning Competition!

On this repo I will share my implementation to this challenge

Mi first competition 😄

Why predict photo popularity?

AI / Machine Learning is the key to the future

A successful solution to this challenge could open doors to further interesting applications of this kind of technology. Imagine smart camera apps that could suggest compositions on the fly or powerful filtering in your social media feeds based on a deeper understanding the kind of visual content you like or dislike. Imagine going on a vacation and a drone follows you about taking photos for you based on the types of compositions, angles and lighting styles that it knows you like, arranging them neatly in your camera roll for you to upload later.

My Implementation

Data Analysis

1. Analyze given Data (JS)
   • Here I decide to use Watson Visual Recognition API, from IBM Watson Developer Cloud
2. Separate Data per User (JS)
3. Add Watson Classes (JS)
4. Generate Train Data (JS)
   • Prepare CSV for Python Implementation

Train Model

Python, Training Multiple Linear Regression Model

Live Predictions

Python, Live Prediction, used trained model.

Model

• Design
• Possible Improves
  • If someone wants to continue, here some ideas 😄

Thanks

Thanks Unpossibly for the oportunity 😄

Innovation Lab UG @InnovationLabUG Innovation Lab UG

Author

Bonnie Milián @BonnieMilianB