Training Latent Diffusion Model
Kernel Inception Distance
The kernel inception distance (KID) is a measure of the similarity between two sets of images. It was introduced by Google AI researcher, Tobias Weyand, in his paper "Kernel Inception Distance: A New Distance Measure for Improved Image Quality Assessment". In this paper, Weyand proposed using the KID to measure the difference between the activations of two sets of images when passed through an inception network, which is a type of deep learning neural network. The KID is calculated as the squared L2 distance between the distributions of the activations of the two sets of images. It is typically used in the field of computer vision to evaluate the performance of image generation algorithms.
Sources
- https://keras.io/examples/generative/ddim #denoising-diffusion-implicit-models
- https://benanne.github.io/2022/01/31/diffusion.html
- https://github.com/apapiu/guided-diffusion-keras
- https://www.louisbouchard.ai/latent-diffusion-models/
- https://www.kaggle.com/code/apapiu/train-latent-diffusion-in-keras-from-scratch
- https://github.com/CompVis/taming-transformers
- https://github.com/CompVis/latent-diffusion
- https://github.com/huggingface/diffusers