Conference Video: Tips and Tricks for Developing Deep Learning Image Models for Production

This is Mengyuan Liu and Ivana Milovanovic’s presentation from the WiDS Puget Sound Conference 2021. Enjoy!

ABSTRACT:

Deep learning models let us extract complex information from images, but can be difficult to deploy in production due to their complexity and latency. This talk covers key considerations and practical tips for preparing deep-learning image models for deployment to server machines and smartphones.

MENGYUAN LIU’S BIO:

Mengyuan is a Senior Data Scientist at SAP Concur working on building AI solutions to revolutionize expense management and reporting. Before joining Concur, Mengyuan obtained a PhD in bioengineering from the University of Washington, specializing in applying machine learning and computer vision technologies to medical imaging. Mengyuan is passionate about supporting women in technology and has been an Ambassador and Lead Organizer for the Women in Data Science Puget Sound Conference for 2 years. Outside work, she is also enthusiastic about environmental issues, ethical fashion, animal welfare and enjoys traveling and good food.

IVANA MILOVANOVIC’S BIO:

Ivana Milovanovic is a data scientist at SAP Concur, where she works on building deep learning algorithms for image and text processing. She holds a PhD in Biomedical Engineering from University of Belgrade, Serbia. Prior to joining SAP Concur, she was a post-doc at University of Washington working in Computational Neuroscience. Her research focused on exploring brain-computer interfaces (BCIs). In her spare time Ivana enjoys hiking, skiing and playing tennis.