RGB-X Model Development: Exploring Four Channel ML Workflows

Machine Learning is rapidly becoming multimodal. With many models in Computer Vision expanding to areas like vision and 3D, one area that has also quietly been advancing rapidly is RGB-X data, such as infrared, depth, or normals. In this talk we will cover some of the leading models in this exploding field of Visual AI and show some best practices on how to work with these complex data formats!

1. RGB-X Model Development
Exploring 4 Channel Model Development
Daniel Gural
Machine Learning & Developer
Relations - Voxel51

2. While Stagnation Fears Rise, Visual AI Marches On



NVDA

3. The Year of Scaling
> Connecting to cloud
> Securing GPUs
> Building out your stack

4. RGB-X Models

5. What is a RGB-X Model?
> A model designed to input or output four channel
images
> Utilized across all industries

6. Why Does This Work Matter?
> Take a step back
> Does it matter today to answer?
> Hold this question for later

7. How RGB-X Models Are Being Used Today

8. Tracking Across all Signals
> Extension of a traditional Object Detection Model
> Tracks objects across frames of any RGB-X image

9. Surveying Difficult Terrain
> Scan hard to access places
> Create a map of areas just by flying through a drone
> Can be performed all onboard

10. Sapiens and Beyond

11. How to Get Started

12. Revisiting the Why
> In a vacuum, none of these use cases move the needle
> So why is this work important?

13. Looking to the Future

14. Looking to the Future

15. Looking to the Future

16. What Does This Mean for Machine Learning?
1.

17. What Does This Mean for Machine Learning?
2.

18. What Does This Mean for Machine Learning?
3.

19. Try Your Hand at Monocular Depth Estimation
https://docs.voxel51.com/tutorials/monocular_depth_estimation.html

20. voxel51.com/jobs/
We are
hiring!