Skip to content
Snippets Groups Projects

Repository graph

You can move around the graph by using the arrow keys.
Select Git revision
0 results
Created with Raphaël 2.2.013Aug10231Jul3022212018remove special character from lhq sample pngmainmainAdded missing samples to READMEadd UMAP for AFHQ CLIPadd UMAP for AFHQ InceptionV3fix SSIM errorupdate main.pymerge class conditional and inpainting evaluation pipelinesadd evaluation helpers and updated readmeupdate evaluation pipelinesolved minor bugs, changes to readme, updated sampling functions for readabilityCombined both versions of the conditional diffusion model, now allows for classifier guided diffusion class conditional, and inpainting image generation.update READMEclipping added to combat class-free guidence lerpmissing dataloader in mainin-out-paintingin-out-paintingfixed UNet erroradded original train and test splitcommit to switch branchesAdded inpainting dataloader (randomly draws black rectangles for masks), concatenation conditioning mechanism for UNet, removed classifier-free guided diffusionchanges to DM class, dataloader, etc.Completed Class Conditional Diffusion Model. This version is intended to be trained on a class labeled dataset. It makes use of classifier-free guided diffusion to boost sample quality of the model when generating images from each class. The conditioning mechanism in the UNet (implemented together with Roy) simply adds an embedding of the class to the time embedding before passing them through the learnable reshaping layers for each block. Successfully trained on 3 class dataset, dogs, cats and wildlife.Initial commit