Pitfalls of Emotion Detection in Production

Justin Shenk (~justin76)


Description:

Deep learning provides many opportunities for businesses to easily scale technology which would have otherwise required thousands of hours of labelling. Using the FER2013 dataset, emotion detection was developed with Peltarion's deep neural network model builder (Deep Emotion, https://github.com/justinshenk/deepemotion). It was implemented as an API with both Keras and the Peltarion API. Some of the challenges in developing this and putting it into production are discussed.

Basic Outline of the Talk

  1. Use cases for emotion detection [4-5 minutes]
  2. Deep learning for facial expression recognition [4-5 minutes]
  3. Training a deep neural network [12-15 minutes]
  4. Brief introduction to convolutional networks
  5. Neural architecture and loss functions
  6. Optimising input data quality
  7. Web deployment and infrastructure [4-5 minutes]
  8. Q/A - [2 minutes]

Who is this talk for?

  • Deep Learning engineers and researchers
  • Python Full Stack developers and data scientists
  • Curious people interested in AI applications

Prerequisites:

  • Basic scripting in Python
  • Interest in machine learning

Speaker Info:

Justin is a Data Scientist at Peltarion and a freelancer for data science solutions.

He developed and demoed computer vision and machine learning applications as an independent developer for Intel at the top conferences (NIPS, CVPR, CEBIT, ICML, PyData, ML Prague) and presented his research at meetups across US, Europe and Asia.

Previously, he started a crowd-sourced translation company.

Speaker Links:

LinkedIn

GitHub

Id: 1167
Section: Data Science, Machine Learning and AI
Type: Talks
Target Audience: Intermediate
Last Updated: