Adversarial Supervision Makes Layout-to-Image Diffusion Models Thrive (ICLR 2024)

Project Page | ArXiv | Code

This model repo contains checkpoints trained on Cityscapes and ADE20K datasets using methods proposed in ALDM. For usage instructions, please refer to our Github.

Model information

ade20k_step9.ckpt and cityscapes_step9.ckpt are pretrained diffusion model weights for inference.

encoder_epoch_50.pth, decoder_epoch_50_20cls.pth and decoder_epoch_50_151cls.pth are segmentation models used for discriminator intialization in training, which are adopted from pretrained UperNet101 here.