Image inpainting tool powered by SOTA AI Model. Remove any unwanted object, defect, people from your pictures or erase and replace(powered by stable diffusion) any thing on your pictures.
Go to file
2022-09-30 22:44:03 +08:00
.github Create FUNDING.yml 2022-06-10 17:01:33 +08:00
assets update README 2022-09-22 22:53:20 +08:00
lama_cleaner enable text_encoder cpu 2022-09-30 22:44:03 +08:00
.gitignore update 2022-04-18 22:18:48 +08:00
Dockerfile Added Dockerfile 2021-11-15 20:11:46 +01:00
LICENSE init 2021-11-15 22:21:01 +08:00
main.py add setup.py 2022-04-18 15:15:51 +08:00
publish.sh update 2022-04-18 22:18:48 +08:00
README.md add run-sd-local arg 2022-09-29 13:13:09 +08:00
requirements-dev.txt add setup.py 2022-04-18 15:15:51 +08:00
requirements.txt enable text_encoder cpu 2022-09-30 22:44:03 +08:00
setup.py 0.20.1 2022-09-27 22:38:28 +08:00

Lama Cleaner

A free and open-source inpainting tool powered by SOTA AI model.

total download version Open in Colab Sanster

img

Features

  • Completely free and open-source
  • Fully self-hosted
  • Classical image inpainting algorithm powered by cv2
  • Multiple SOTA AI models
    1. LaMa
    2. LDM
    3. ZITS
    4. MAT
    5. FcF
    6. SD1.4
  • Support CPU & GPU
  • Various inpainting strategy
  • Run as a desktop APP

Usage

Usage Before After
Remove unwanted things unwant_object2 unwant_object2
Remove unwanted person unwant_person unwant_person
Remove Text text text
Remove watermark watermark watermark_clean
Fix old photo oldphoto oldphoto_clean
Text Driven Inpainting dog Prompt: a fox sitting on a bench
fox

Quick Start

pip install lama-cleaner

# Model will be downloaded automatically
lama-cleaner --model=lama --device=cpu --port=8080
# Lama Cleaner is now running at http://localhost:8080

Available arguments:

Name Description Default
--model lama/ldm/zits/mat/fcf/sd1.4 See details in Inpaint Model lama
--hf_access_token stable-diffusion(sd) model need huggingface access token to download model
--sd-run-local Once the model as downloaded, you can pass this arg and remove --hf_access_token
--sd-disable-nsfw Disable stable-diffusion NSFW checker.
--sd-cpu-textencoder Always run stable-diffusion TextEncoder model on CPU.
--device cuda or cpu cuda
--port Port for backend flask web server 8080
--gui Launch lama-cleaner as a desktop application
--gui_size Set the window size for the application 1200 900
--input Path to image you want to load by default None
--debug Enable debug mode for flask web server

Inpainting Model

Model Description Config
cv2 👍 No GPU is required, and for simple backgrounds, the results may even be better than AI models.
LaMa 👍 Generalizes well on high resolutions(~2k)
LDM 👍 Possible to get better and more detail result
👍 The balance of time and quality can be achieved by adjusting steps
😐 Slower than GAN model
😐 Need more GPU memory
Steps: You can get better result with large steps, but it will be more time-consuming
Sampler: ddim or plms. In general plms can get better results with fewer steps
ZITS 👍 Better holistic structures compared with previous methods
😐 Wireframe module is very slow on CPU
Wireframe: Enable edge and line detect
MAT TODO
FcF 👍 Better structure and texture generation
😐 Only support fixed size (512x512) input
SD1.4 👍 SOTA text-to-image diffusion model

LaMa vs LDM

Original Image LaMa LDM
photo-1583445095369-9c651e7e5d34 photo-1583445095369-9c651e7e5d34_cleanup_lama photo-1583445095369-9c651e7e5d34_cleanup_ldm

LaMa vs ZITS

Original Image ZITS LaMa
zits_original zits_compare_zits zits_compare_lama

Image is from ZITS paper. I didn't find a good example to show the advantages of ZITS and let me know if you have a good example. There can also be possible problems with my code, if you find them, please let me know too!

LaMa vs FcF

Original Image Lama FcF
texture texture_lama texture_fcf

Inpainting Strategy

Lama Cleaner provides three ways to run inpainting model on images, you can change it in the settings dialog.

Strategy Description VRAM
Original Use the resolution of the original image 🎉
Resize Resize the image to a smaller size before inpainting. Lama Cleaner will make sure that the area of the image outside the mask is not degraded. 🎉 🎉
Crop Crop masking area from the original image to do inpainting 🎉 🎉 🎉

Download Model Mannually

If you have problems downloading the model automatically when lama-cleaner start, you can download it manually. By default lama-cleaner will load model from TORCH_HOME=~/.cache/torch/hub/checkpoints/, you can set TORCH_HOME to other folder and put the models there.

Development

Only needed if you plan to modify the frontend and recompile yourself.

Frontend

Frontend code are modified from cleanup.pictures, You can experience their great online services here.

  • Install dependencies:cd lama_cleaner/app/ && yarn
  • Start development server: yarn start
  • Build: yarn build

Docker

Run within a Docker container. Set the CACHE_DIR to models location path. Optionally add a -d option to the docker run command below to run as a daemon.

Build Docker image

docker build -f Dockerfile -t lamacleaner .

Run Docker (cpu)

docker run -p 8080:8080 -e CACHE_DIR=/app/models -v  $(pwd)/models:/app/models -v $(pwd):/app --rm lamacleaner python3 main.py --device=cpu --port=8080

Run Docker (gpu)

docker run --gpus all -p 8080:8080 -e CACHE_DIR=/app/models -v $(pwd)/models:/app/models -v $(pwd):/app --rm lamacleaner python3 main.py --device=cuda --port=8080

Then open http://localhost:8080