Deep learning based image and Image/Text Generative models

Denoising Diffusion Implicit Models

On Oxford Flower dataset Image size 128 x 128

From noise to final images in 200 steps

Final Images @200 step

After training basic diffusion model on flower dataset we will now train the ImageGen diffusion model on small splash dataset with intermediate resolution

ImageGen

After training ImageGen on unconditional small splash dataset, novel images generated of intermediate resolution (512 x 512) by imagegen looks like below

       

       

       

       

Code and trained model will bre realsed soon stay tuned!!

References

1) @article{Song, Jiaming, Chenlin Meng, and Stefano Ermon. "Denoising diffusion implicit models." arXiv preprint arXiv:2010.02502 (2020)}
2) https://imagen.research.google/
3) https://github.com/lucidrains/imagen-pytorch