Embedded Deep Learning Super Resolution on GStreamer Using ONNX Inference Runtime

Explore embedded deep learning super resolution techniques using GStreamer and ONNX Inference Runtime in this 27-minute conference talk presented by Aaron Boxer and Marcus Edel from Collabora Inc. Delve into the importance of upscaling in image processing and understand the concept of super resolution, including blind super resolution. Examine various types of super resolution methods and learn about kernel estimation. Gain insights into the qualitative and quantitative results of these techniques, and discover how to implement them using GStreamer. This talk provides a comprehensive overview of super resolution in embedded systems, offering valuable knowledge for developers and researchers working on image enhancement applications.

Intro
Super-Resolution
Why (up)scaling matters
Solved Problem
What is Super Resolution?
Blind Super Resolution?
Super Resolution types
Kernel Estimation
Overview
Qualitative Results
Quantitative Results
GStreamer Implementation