Glow: Better reversible generative models
Glow is a new type of generative model that uses invertible 1x1 convolutions, making it reversible. This allows it to generate high-resolution images efficiently and discover features that can be used to manipulate data attributes. The model builds on previous work in the field and simplifies its architecture. OpenAI has released code for Glow and an online visualization tool so researchers can explore and build upon these results.
Glow is a new type of generative model that uses invertible 1x1 convolutions, making it reversible. This allows it to generate high-resolution images efficiently and discover features that can be used to manipulate data attributes. The model builds on previous work in the field and simplifies its architecture. OpenAI has released code for Glow and an online visualization tool so researchers can explore and build upon these results.
---
Why it matters: This matters because Glow's efficiency and ability to generate high-resolution images could improve the performance of various AI applications, such as image recognition and generation tasks.
Source: https://openai.com/index/glow
This article was originally published at: https://openai.com/index/glow