🦙 LaMa: Resolution-robust Large Mask Inpainting with Fourier Convolutions

Official implementation by Samsung Research

by Roman Suvorov, Elizaveta Logacheva, Anton Mashikhin, Anastasia Remizova, Arsenii Ashukha, Aleksei Silvestrov, Naejin Kong, Harshith Goka, Kiwoong Park, Victor Lempitsky.

🔥 🔥 🔥
LaMa generalizes surprisingly well to much higher resolutions (~2k ❗️ ) than it saw during training (256x256), and achieves the excellent performance even in challenging scenarios, e.g. completion of periodic structures.

[Project page] [arXiv] [Supplementary] [BibTeX]

Try out in Google Colab

Environment setup

Clone the repo: git clone https://github.com/saic-mdal/lama.git

There are three options of an environment:

Python virtualenv:

virtualenv inpenv --python=/usr/bin/python3
source inpenv/bin/activate
pip install torch==1.8.0 torchvision==0.9.0

cd lama
pip install -r requirements.txt

Conda

% Install conda for Linux, for other OS download miniconda at https://docs.conda.io/en/latest/miniconda.html
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
bash Miniconda3-latest-Linux-x86_64.sh -b -p $HOME/miniconda
$HOME/miniconda/bin/conda init bash

cd lama
conda env create -f conda_env.yml
conda activate lama
conda install pytorch torchvision torchaudio cudatoolkit=10.2 -c pytorch -y
pip install pytorch-lightning==1.2.9

Docker: No actions are needed 🎉 .

Inference

Run

cd lama
export TORCH_HOME=$(pwd) && export PYTHONPATH=.

1. Download pre-trained models

Install tool for yandex disk link extraction:

pip3 install wldhx.yadisk-direct

The best model (Places2, Places Challenge):

curl -L $(yadisk-direct https://disk.yandex.ru/d/ouP6l8VJ0HpMZg) -o big-lama.zip
unzip big-lama.zip

All models (Places & CelebA-HQ):

curl -L $(yadisk-direct https://disk.yandex.ru/d/EgqaSnLohjuzAg) -o lama-models.zip
unzip lama-models.zip

2. Prepare images and masks

Download test images:

curl -L $(yadisk-direct https://disk.yandex.ru/d/xKQJZeVRk5vLlQ) -o LaMa_test_images.zip
unzip LaMa_test_images.zip

OR prepare your data:

1) Create masks named as `[images_name]_maskXXX[image_suffix]`, put images and masks in the same folder.

You can use the script for random masks generation.

Check the format of the files:

image1_mask001.png
image1.png
image2_mask001.png
image2.png

Specify image_suffix, e.g. .png or .jpg or _input.jpg in configs/prediction/default.yaml.

3. Predict

On the host machine:

python3 bin/predict.py model.path=$(pwd)/big-lama indir=$(pwd)/LaMa_test_images outdir=$(pwd)/output

OR in the docker

The following command will pull the docker image from Docker Hub and execute the prediction script

bash docker/2_predict.sh $(pwd)/big-lama $(pwd)/LaMa_test_images $(pwd)/output device=cpu

Docker cuda: TODO

Train and Eval

⚠️ Warning: The training is not fully tested yet, e.g., did not re-training after refactoring ⚠️

Make sure you run:

cd lama
export TORCH_HOME=$(pwd) && export PYTHONPATH=.

Then download models for perceptual loss:

mkdir -p ade20k/ade20k-resnet50dilated-ppm_deepsup/
wget -P ade20k/ade20k-resnet50dilated-ppm_deepsup/ http://sceneparsing.csail.mit.edu/model/pytorch/ade20k-resnet50dilated-ppm_deepsup/encoder_epoch_20.pth

Places

On the host machine:

# Download data from http://places2.csail.mit.edu/download.html
# Places365-Standard: Train(105GB)/Test(19GB)/Val(2.1GB) from High-resolution images section
wget http://data.csail.mit.edu/places/places365/train_large_places365standard.tar
wget http://data.csail.mit.edu/places/places365/val_large.tar
wget http://data.csail.mit.edu/places/places365/test_large.tar

# Unpack and etc.
bash fetch_data/places_standard_train_prepare.sh
bash fetch_data/places_standard_test_val_prepare.sh
bash fetch_data/places_standard_evaluation_prepare_data.sh

# Sample images for test and viz at the end of epoch
bash fetch_data/places_standard_test_val_sample.sh
bash fetch_data/places_standard_test_val_gen_masks.sh

# Run training
# You can change bs with data.batch_size=10
python bin/train.py -cn lama-fourier location=places_standard

# Infer model on thick/thin/medium masks in 256 and 512 and run evaluation 
# like this:
python3 bin/predict.py \
model.path=$(pwd)/experiments/
   
    _
    
     _lama-fourier_/ \
indir=$(pwd)/places_standard_dataset/evaluation/random_thick_512/ \
outdir=$(pwd)/inference/random_thick_512 model.checkpoint=last.ckpt

python3 bin/evaluate_predicts.py \
$(pwd)/configs/eval_2gpu.yaml \
$(pwd)/places_standard_dataset/evaluation/random_thick_512/ \
$(pwd)/inference/random_thick_512 $(pwd)/inference/random_thick_512_metrics.csv

Docker: TODO

CelebA

On the host machine:

# Make shure you are in lama folder
cd lama
export TORCH_HOME=$(pwd) && export PYTHONPATH=.

# Download CelebA-HQ dataset
# Download data256x256.zip from https://drive.google.com/drive/folders/11Vz0fqHS2rXDb5pprgTjpD7S2BAJhi1P

# unzip & split into train/test/visualization & create config for it
bash fetch_data/celebahq_dataset_prepare.sh

# generate masks for test and visual_test at the end of epoch
bash fetch_data/celebahq_gen_masks.sh

# Run training
python bin/train.py -cn lama-fourier-celeba data.batch_size=10

# Infer model on thick/thin/medium masks in 256 and run evaluation 
# like this:
python3 bin/predict.py \
model.path=$(pwd)/experiments/
   
    _
    
     _lama-fourier-celeba_/ \
indir=$(pwd)/celeba-hq-dataset/visual_test_256/random_thick_256/ \
outdir=$(pwd)/inference/celeba_random_thick_256 model.checkpoint=last.ckpt

Docker: TODO

Places Challenge

On the host machine:

# This script downloads multiple .tar files in parallel and unpacks them
# Places365-Challenge: Train(476GB) from High-resolution images (to train Big-Lama) 
bash places_challenge_train_download.sh

TODO: prepare
TODO: train 
TODO: eval

Docker: TODO

Create your data

On the host machine:

Explain explain explain

TODO: format
TODO: configs 
TODO: run training
TODO: run eval

OR in the docker:

TODO: train
TODO: eval

Hints

Generate different kinds of masks

The following command will execute a script that generates random masks.

bash docker/1_generate_masks_from_raw_images.sh \
    configs/data_gen/random_medium_512.yaml \
    /directory_with_input_images \
    /directory_where_to_store_images_and_masks \
    --ext png

The test data generation command stores images in the format, which is suitable for prediction.

The table below describes which configs we used to generate different test sets from the paper. Note that we do not fix a random seed, so the results will be slightly different each time.

	Places 512x512	CelebA 256x256
Narrow	random_thin_512.yaml	random_thin_256.yaml
Medium	random_medium_512.yaml	random_medium_256.yaml
Wide	random_thick_512.yaml	random_thick_256.yaml

Feel free to change the config path (argument #1) to any other config in configs/data_gen or adjust config files themselves.

Override parameters in configs

Also you can override parameters in config like this:

python3 bin/train.py -cn 
   
     data.batch_size=10 run_title=my-title

Where .yaml file extension is omitted

Models options

Config names for models from paper (substitude into the training command):

* big-lama
* big-lama-regular
* lama-fourier
* lama-regular
* lama_small_train_masks

Which are seated in configs/training/folder

Training time & resources

TODO

Acknowledgments

Segmentation code and models if form CSAILVision.
LPIPS metric is from richzhang
SSIM is from Po-Hsun-Su
FID is from mseitzer

Citation

If you found this code helpful, please consider citing:

@article{suvorov2021resolution,
  title={Resolution-robust Large Mask Inpainting with Fourier Convolutions},
  author={Suvorov, Roman and Logacheva, Elizaveta and Mashikhin, Anton and Remizova, Anastasia and Ashukha, Arsenii and Silvestrov, Aleksei and Kong, Naejin and Goka, Harshith and Park, Kiwoong and Lempitsky, Victor},
  journal={arXiv preprint arXiv:2109.07161},
  year={2021}
}

🦙 LaMa Image Inpainting, Resolution-robust Large Mask Inpainting with Fourier Convolutions, WACV 2022

Related tags

Overview

🦙 LaMa: Resolution-robust Large Mask Inpainting with Fourier Convolutions

Environment setup

Inference

Train and Eval

Places

CelebA

Places Challenge

Create your data

Hints

Generate different kinds of masks

Override parameters in configs

Models options

Links

Training time & resources

Acknowledgments

Citation

Owner

Advanced Image Manipulation Lab @ Samsung AI Center Moscow

Ipython notebook presentations for getting starting with basic programming, statistics and machine learning techniques

Data loaders and abstractions for text and NLP

CLIP (Contrastive Language–Image Pre-training) for Italian

Architecture Patterns with Python (TDD, DDD, EDM)

Replication Code for "Self-Supervised Bug Detection and Repair" NeurIPS 2021

A PyTorch implementation of "Predict then Propagate: Graph Neural Networks meet Personalized PageRank" (ICLR 2019).

OptaPlanner wrappers for Python. Currently significantly slower than OptaPlanner in Java or Kotlin.

ToFFi - Toolbox for Frequency-based Fingerprinting of Brain Signals

Attention-driven Robot Manipulation (ARM) which includes Q-attention

A deep learning network built with TensorFlow and Keras to classify gender and estimate age.

CFC-Net: A Critical Feature Capturing Network for Arbitrary-Oriented Object Detection in Remote Sensing Images

CHERRY is a python library for predicting the interactions between viral and prokaryotic genomes

Google Landmark Recogntion and Retrieval 2021 Solutions

A style-based Quantum Generative Adversarial Network

Script for getting information in discord

RLMeta is a light-weight flexible framework for Distributed Reinforcement Learning Research.

Interactive Image Segmentation via Backpropagating Refinement Scheme

Implementation of SETR model, Original paper: Rethinking Semantic Segmentation from a Sequence-to-Sequence Perspective with Transformers.

Resilient projection-based consensus actor-critic (RPBCAC) algorithm

Python implementation of cover trees, near-drop-in replacement for scipy.spatial.kdtree