ControlNet-different backbones

In this fork of the original ControlNet repo, we evaluated the possibility of using lighter backbones for ControlNet model instead of Stable Diffusion encoder. Our experiments and findings are described here: https://medium.com/@dzvinkayarish/controlnetlite-smaller-and-faster-controlnet-55b26cef946e . The project was inspired by this discussion lllyasviel/ControlNet#188 .

Name		Name	Last commit message	Last commit date
Latest commit History 1 Commit
annotator		annotator
cldm		cldm
docs		docs
fid		fid
font		font
github_page		github_page
ldm		ldm
models		models
notebooks		notebooks
test_imgs		test_imgs
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
aesthetics_predictor.py		aesthetics_predictor.py
config.py		config.py
environment.yaml		environment.yaml
get_val_data_cc3m.py		get_val_data_cc3m.py
get_val_data_laion.py		get_val_data_laion.py
get_val_data_things.py		get_val_data_things.py
gif_creator.py		gif_creator.py
gradio_annotator.py		gradio_annotator.py
gradio_canny2image.py		gradio_canny2image.py
gradio_depth2image.py		gradio_depth2image.py
gradio_fake_scribble2image.py		gradio_fake_scribble2image.py
gradio_hed2image.py		gradio_hed2image.py
gradio_hough2image.py		gradio_hough2image.py
gradio_normal2image.py		gradio_normal2image.py
gradio_pose2image.py		gradio_pose2image.py
gradio_scribble2image.py		gradio_scribble2image.py
gradio_scribble2image_interactive.py		gradio_scribble2image_interactive.py
gradio_seg2image.py		gradio_seg2image.py
laion_dataset.py		laion_dataset.py
requirements_local.txt		requirements_local.txt
share.py		share.py
tool_add_control.py		tool_add_control.py
tool_add_control_sd21.py		tool_add_control_sd21.py
tool_transfer_control.py		tool_transfer_control.py
train.sh		train.sh
tutorial_dataset.py		tutorial_dataset.py
tutorial_dataset_test.py		tutorial_dataset_test.py
tutorial_train.py		tutorial_train.py
tutorial_train_sd21.py		tutorial_train_sd21.py

Provide feedback