LaCon: Late-Constraint Diffusion for Steerable Guided Image Synthesis
Chang Liu, Rui Li, Kaidong Zhang, Xin Luo, Dong Liu
[Paper
] / [Project
] / [GitHub
] / [ModelScope
] / [Demo
]
Official pre-trained model weights of the paper titled "LaCon: Late-Constraint Diffusion for Steerable Guided Image Synthesis".
Model Weights Structure
LaCon/
βββ data-preprocessing
βββ bdcn.pth <----- Model weights of BDCN edge detector (Place in `data-preprocessing/bdcn-edge-detection/checkpoints`)
βββ vgg16.pth <----- Preresquisites of BDCN edge detector (Place in `data-preprocessing/bdcn-edge-detection/checkpoints`)
βββ u2net.pth <----- Model weights of U2-Net saliency detector (Place in `data-preprocessing/u2net-saliency-detection/checkpoints`)
βββ sd-checkpoints
βββ sd-v1-4.ckpt <----- SD v1.4
βββ kl-f8.ckpt <----- VAE model of SD v1.4
βββ sd-celebahq.ckpt <----- SD CelebA-HQ
βββ vq-f4.ckpt <----- VAE model of SD CelebA-HQ
βββ sdv14_edge.pth <----- Condition aligner based on SD v1.4, for edge conditions
βββ sdv14_color.pth <----- Condition aligner based on SD v1.4, for color conditions
βββ sdv14_mask.pth <----- Condition aligner based on SD v1.4, for mask conditions
βββ sd_celeb_edge.pth <----- Condition aligner based on SD CelebA-HQ, for edge conditions
βββ sd_celeb_stroke.pth <----- Condition aligner based on SD CelebA-HQ, for color conditions