Views
No views yet
[!WARNING] we do not have a full checkpoint conversion validation, if you encounter pipeline loading failure and unsidered output, please contact me via bili_sakura@zju.edu.cn
This is an image generation model only, not a change-detection pipeline. For change detection, use this UNet as a feature extractor and add a CD head (see pretrained-cd-models).
custom_pipeline (pipeline.py is in the repo, use relative path):1from diffusers import DDIMScheduler, DiffusionPipeline
2
3pipe = DiffusionPipeline.from_pretrained(
4 "BiliSakura/ddpm-cd-pretrained-256",
5 custom_pipeline="pipeline",
6 trust_remote_code=True,
7).to("cuda")
8pipe.scheduler = DDIMScheduler.from_config(pipe.scheduler.config)
9
10# Control speed vs quality with num_inference_steps (default 2000). Use fewer (e.g. 50–250) for faster inference.
11images = pipe.generate(batch_size=4, image_size=256, num_inference_steps=50)train_cd.py; pre-trained CD heads are in ddpm-cd.1accelerate launch scripts/train_cd.py \
2 --pretrained_model_path BiliSakura/ddpm-cd-pretrained-256 \
3 --train_data_dir dataset/LEVIR-CD256 \
4 --val_data_dir dataset/LEVIR-CD256 \
5 --output_dir experiments/cd-levir \
6 --resolution 256 \
7 --timesteps 50 100 400 \
8 --feat_type dec1@inproceedings{bandaraDDPMCDDenoisingDiffusion2025,
2 title = {{{DDPM-CD}}: {{Denoising Diffusion Probabilistic Models}} as {{Feature Extractors}} for {{Remote Sensing Change Detection}}},
3 shorttitle = {{{DDPM-CD}}},
4 booktitle = {Proceedings of the {{Winter Conference}} on {{Applications}} of {{Computer Vision}}},
5 author = {Bandara, Wele Gedara Chaminda and Nair, Nithin Gopalakrishnan and Patel, Vishal},
6 year = 2025,
7 pages = {5250--5262},
8 urldate = {2025-12-28}
9}