Xpression

Project Xpression: A winning hackathon project I worked with a team on to win The Teradata Labs Artificial Intelligence and Cognitive Services Hackathon 2017 as a freshman at UC San Diego. This is my personal copy. I specialized on the implementation of Aster analytics software. I created two classification models using two different techniques, General Linear Model and A Support Vector Machine. Once we identified that the SVM yield the most accurate results, we used that to classify the drivers into one of ten categories of distraction levels.

---

         _   _ _____ _____ _____ _____ _____ _____ _____ _____              
        \ \ / /  _  |  _  \  ___|  ___|  ___|_   _|  _  |  _  |             
         \ v /| |_| | |_| / |___| |___| |___  | | | | | | | | |             
          | | |  ___|  _  \  ___|___  |___  | | | | | | | | | |             
         / ^ \| |   | | | | |___ ___| |___| |_| |_| |_| | | | |             
        /_/ \_\_|   |_| |_|_____|_____|_____|_____|_____|_| |_|             

---

Machine Learning application using AWS Rekognition and Teradata Aster used to classify drivers into one of ten categories, each representing different levels of distraction.

Dependencies:

• AWSCLI
  • Access to AWS Rekognition
• Teradata Services
  • Access to Teradata Aster
  • Access to TDBMS

---

Program Overview: There are several portions of the pipeline that need to be run in order to preprocess data, store it, and then operate on it using Aster.

1. Preprocessing

• Ensure AWSCLI is set up with the proper credentials.
• Run utility.py's functions as appropriate. It uses RekognitionInterface's definitions to use pass images through AWS Rekognition. The returned results can be placed in a CSV file for easy importing into a Teradata database.

2. Storing data

• Once the database is set up, run TestImages.py to place the data into the database using a SQL query.

3. Using Aster

• Use the SVM function on the imported data to classify image features.

---