Creative Application of Machine learning for Digital Elevation Models
Activity: Talk or presentation types › Invited talk
Dongjoe Shin (Speaker), 23 Oct 2019
The spatial resolution of a Digital Elevation Model (DEM) plays a crucial role in many practical remote sensing applications. However, it is normally fixed to the sensor specification and its operating conditions (e.g. the resolution of a spaceborne sensor is normally lower than that of an airborne sensor).
One interesting idea is “can we make a machine to figure out the higher resolution terrain model from its low-resolution counterpart?” It sounds a bit absurd, but humans are doing this data interpolation seamlessly every day; we can figure out the details of a local geometry from smoothed data or even guess the missing values from an occluded scene.
To solve this problem, I built a deep learning model motivated by the latest super resolution models and trained it with a large number of sample height maps around the world. This talk gave a short overview of the outcomes, and discussed the future opportunities in creative applications and HCI
One interesting idea is “can we make a machine to figure out the higher resolution terrain model from its low-resolution counterpart?” It sounds a bit absurd, but humans are doing this data interpolation seamlessly every day; we can figure out the details of a local geometry from smoothed data or even guess the missing values from an occluded scene.
To solve this problem, I built a deep learning model motivated by the latest super resolution models and trained it with a large number of sample height maps around the world. This talk gave a short overview of the outcomes, and discussed the future opportunities in creative applications and HCI
23 Oct 2019
School
| Name | School of Creative Technologies |
|---|---|
| Period | 7/07/00 → … |
| Location | Winston Churchill Avenue |
| City | Portsmouth |
| Country | United Kingdom |
Related information
ID: 16057063
