This is an Implementation for Denoising Diffusion Implicit Models (DDIM)

DDIM is one of the denoising diffusion probabilistic models family but the key difference here it doesn't require a large reverse diffusion time steps to produce the samples or images as you can see from above this gif was created with 25 reverse diffusion time steps.

---

if you want to train your own model for a specific dataset this colab is for you.

if you want to try the pretrained model this colab is for you

Content:

• An implementation of Denoising Diffusion Implicit Models (DDIM) with continuous time. All variables are properly named and the code is densely commented.
• A pretrained weights for a model trained on Flowers Dataset from the university of oxford .
• A Diffusion model combining all the important parts to generate samples
• Some helper methods for visualizing the generated samples

---

Code to try the pretrained model:

from Diffusion import DiffusionModel, DiffUnet 
import matplotlib.pyplot as plt

net = DiffUnet(block_depth=2) # the Unet model for the diffusion module
model = DiffusionModel(net, num_steps=1000, input_res=[64,64])  # initialize the module with 1000 training steps and img size (64, 64)

model.load("Pretrained") # load the pretrained weights from the weights directory
sample = model.generate(num_samples=1, num_infer_steps=25).cpu().numpy().squeeze() # generate one sample with 25 reverse diffusion steps

sample = model.inverse_transform(sample) # reverse the transformation to better visualize the image

# visualize the image
plt.imshow(sample)
plt.axis("off");