Image GPT

📰 OpenAI News

Image GPT generates coherent image completions and samples using a transformer model trained on pixel sequences

advanced Published 17 Jun 2020
Action Steps
  1. Train a large transformer model on pixel sequences
  2. Evaluate the model's performance on image completion and sampling tasks
  3. Compare the model's features to those of top convolutional nets in the unsupervised setting
  4. Explore applications of Image GPT in computer vision and graphics design
Who Needs to Know This

AI engineers and researchers can benefit from this breakthrough, as it opens up new possibilities for image generation and processing, and can be applied to various applications such as computer vision and graphics design

Key Insight

💡 Transformer models can be used for image generation and processing, achieving competitive results with convolutional nets

Share This
🖼️ Image GPT generates coherent images using transformer models!

Key Takeaways

Image GPT generates coherent image completions and samples using a transformer model trained on pixel sequences

Full Article

We find that, just as a large transformer model trained on language can generate coherent text, the same exact model trained on pixel sequences can generate coherent image completions and samples. By establishing a correlation between sample quality and image classification accuracy, we show that our best generative model also contains features competitive with top convolutional nets in the unsupervised setting.
Read full article → ← Back to Reads