Welcome to MMEditing’s documentation!

Languages: English | 简体中文

MMEditing is an open-source toolbox for low-level vision. It supports various tasks, including:

• Image super-resolution

• Video super-resolution

• Video frame interpolation

• Image inpainting

• Image matting

• Image generation

• and possibly more in the future.

It is based on PyTorch and is a part of the OpenMMLab project. Codes are available on GitHub.

Documentation

Get Started

• Installation
  • Best Practices
  • Customize Installation
  • Additional Notes
• Demo
  • Inpainting
  • Matting
  • Restoration (Image)
  • Restoration (Face Image)
  • Restoration (Video)
  • video frame interpolation
  • Generation

Basic Usage

• Inference with pre-trained models
  • Test a pre-trained model
• Train a model
  • Train with single/multiple GPUs
  • Train with Slurm
  • Launch multiple jobs on a single machine
• Learn about Configs
  • Config Name Style
  • Config System for Generation
  • Config System for Inpainting
  • Config System for Matting
  • Config System for Restoration
• Use intermediate variables in configs

Customize MMEditing

• Tutorial 1: Customize Datasets
  • Supported Data Format
  • Support new data format
  • Customize datasets by dataset wrappers
• Tutorial 2: Customize Data Pipelines
  • Design of Data pipelines
  • Extend and use custom pipelines
• Tutorial 3: Customize Models
  • Model Argument
  • generator
  • Specification in Configuration File
  • pixel_loss
  • train_cfg and test_cfg
  • Model Functions
• Tutorial 4: Customize Losses
  • Design of loss modules

Model Zoo

• Overview
  • Inpainting Models
  • Matting Models
  • Super-Resolution Models
  • Generation Models
  • Frame-Interpolation Models
• Inpainting Models
  • AOT-GAN (TVCG’2021)
  • DeepFillv1 (CVPR’2018)
  • DeepFillv2 (CVPR’2019)
  • Global&Local (ToG’2017)
  • PConv (ECCV’2018)
• Matting Models
  • DIM (CVPR’2017)
  • GCA (AAAI’2020)
  • IndexNet (ICCV’2019)
• Super-Resolution Models
  • BasicVSR (CVPR’2021)
  • BasicVSR++ (CVPR’2022)
  • DIC (CVPR’2020)
  • EDSR (CVPR’2017)
  • EDVR (CVPRW’2019)
  • ESRGAN (ECCVW’2018)
  • GLEAN (CVPR’2021)
  • IconVSR (CVPR’2021)
  • LIIF (CVPR’2021)
  • RDN (CVPR’2018)
  • RealBasicVSR (CVPR’2022)
  • Real-ESRGAN (ICCVW’2021)
  • SRCNN (TPAMI’2015)
  • SRGAN (CVPR’2016)
  • TDAN (CVPR’2020)
  • TOFlow (IJCV’2019)
  • TTSR (CVPR’2020)
• Generation Models
  • CycleGAN (ICCV’2017)
  • Pix2Pix (CVPR’2017)
• Frame-Interpolation Models
  • CAIN (AAAI’2020)
  • FLAVR (arXiv’2020)
  • TOFlow (IJCV’2019)

Datasets

• Overview
  • Generation Datasets
  • Inpainting Datasets
  • Matting Datasets
  • Super-Resolution Datasets
  • Video Frame Interpolation Datasets
• Inpainting Datasets
  • CelebA-HQ Dataset
  • Paris Street View Dataset
  • Places365 Dataset
• Matting Datasets
  • Composition-1k Dataset
• Super-Resolution Datasets
  • DF2K_OST Dataset
  • DIV2K Dataset
  • REDS Dataset
  • Vid4 Dataset
  • Vimeo90K Dataset
• Generation Datasets
  • Paired Dataset for Pix2pix
  • Unpaired Dataset for CycleGAN

Useful Tools and Scripts

• Useful tools
  • Get the FLOPs and params (experimental)
  • Publish a model
  • Convert to ONNX (experimental)
  • Convert ONNX to TensorRT (experimental)
  • Evaluate ONNX and TensorRT Models (experimental)

API Reference

• mmedit.core
• mmedit.datasets
• mmedit.datasets.pipelines
• mmedit.models
  • models
  • common
  • backbones
  • components
  • losses
• mmedit.utils

Miscellaneous

• Changelog
  • v0.15.1 (04/07/2022)
  • v0.15.0 (01/06/2022)
  • v0.14.0 (01/04/2022)
  • v0.13.0 (01/03/2022)
  • v0.12.0 (31/12/2021)
  • v0.11.0 (03/11/2021)
  • v0.10.0 (12/08/2021).
  • v0.9.0 (30/06/2021).
  • v0.8.0 (31/05/2021).
  • v0.7.0 (30/04/2021).
  • v0.6.0 (08/04/2021).
  • v0.5.0 (09/07/2020).

Switch Language

• English
• 简体中文

Indices and tables

• 索引

• 模块索引

• 搜索页面