Can a color really beat the AI revolution? For now, it looks like it can.
Subscribe and turn on notifications ? so you don't miss any videos: http://goo.gl/0bsAjO
The greenscreen is a staple of visual effects — and it may stick around even in the age of AI “magic.” The video above explains why.
It turns out that greenscreens, while imperfect, provide certain background separation benefits that are tough for AI to replicate due to the way it’s been trained and the limitations of available data. Preparation can help improve results, but this video shows why, ultimately, AI tools will remain one in a suite of options rather than a greenscreen killer.
Further reading
https://www.cs.unc.edu/~ronisen/
You can find more of Soumyadip (Roni) Sengupta’s papers here, including links to his various greenscreen work.
https://segment-anything.com/
If you want to try the latest breakthrough in image segmentation, Meta’s demo lets you upload your own images.
https://runwayml.com/
https://research.runwayml.com/publications/towards-unified-keyframe-propagation-models
Runway ML has a variety of tools that sit on top of AI infrastructure, allowing you to play with all sorts of applications. They’ve documented a bit of their process at the link above.
https://openaccess.thecvf.com/content_cvpr_2017/papers/Xu_Deep_Image_Matting_CVPR_2017_paper.pdf
If you want to really learn more, you can read this paper that explains image matting on an important dataset.
Make sure you never miss behind-the-scenes content in the Vox Video newsletter, sign up here: http://vox.com/video-newsletter
Vox is an explanatory newsroom on a mission to help everyone understand our weird, wonderful, complicated world, so that we can all help shape it. Part of that mission is keeping our work free. You can help us do that by making a gift: http://www.vox.com/contribute-now
Watch our full video catalog: http://goo.gl/IZONyE
Follow Vox on TikTok: http://tiktok.com/@voxdotcom
Check out our articles: https://www.vox.com/
Listen to our podcasts: https://www.vox.com/podcasts