Official PyTorch code for Mutual Affine Network for Spatially Variant Kernel Estimation in Blind Image Super-Resolution (MANet, ICCV2021)

Overview

Mutual Affine Network for Spatially Variant Kernel Estimation in Blind Image Super-Resolution (MANet, ICCV2021)

This repository is the official PyTorch implementation of Mutual Affine Network for Spatially Variant Kernel Estimation in Blind Image Super-Resolution (arxiv, supplementary).

🚀 🚀 🚀 News:


Existing blind image super-resolution (SR) methods mostly assume blur kernels are spatially invariant across the whole image. However, such an assumption is rarely applicable for real images whose blur kernels are usually spatially variant due to factors such as object motion and out-of-focus. Hence, existing blind SR methods would inevitably give rise to poor performance in real applications. To address this issue, this paper proposes a mutual affine network (MANet) for spatially variant kernel estimation. Specifically, MANet has two distinctive features. First, it has a moderate receptive field so as to keep the locality of degradation. Second, it involves a new mutual affine convolution (MAConv) layer that enhances feature expressiveness without increasing receptive field, model size and computation burden. This is made possible through exploiting channel interdependence, which applies each channel split with an affine transformation module whose input are the rest channel splits. Extensive experiments on synthetic and real images show that the proposed MANet not only performs favorably for both spatially variant and invariant kernel estimation, but also leads to state-of-the-art blind SR performance when combined with non-blind SR methods.

Requirements

  • Python 3.7, PyTorch >= 1.6, scipy >= 1.6.3
  • Requirements: opencv-python
  • Platforms: Ubuntu 16.04, cuda-10.0 & cuDNN v-7.5

Note: this repository is based on BasicSR. Please refer to their repository for a better understanding of the code framework.

Quick Run

Download stage3_MANet+RRDB_x4.pth from release and put it in ./pretrained_models. Then, run this command:

cd codes
python test.py --opt options/test/test_stage3.yml

Data Preparation

To prepare data, put training and testing sets in ./datasets as ./datasets/DIV2K/HR/0801.png. Commonly used datasets can be downloaded here.

Training

Step1: to train MANet, run this command:

python train.py --opt options/train/train_stage1.yml

Step2: to train non-blind RRDB, run this command:

python train.py --opt options/train/train_stage2.yml

Step3: to fine-tune RRDB with MANet, run this command:

python train.py --opt options/train/train_stage3.yml

All trained models can be downloaded from release. For testing, downloading stage3 models is enough.

Testing

To test MANet (stage1, kernel estimation only), run this command:

python test.py --opt options/test/test_stage1.yml

To test RRDB-SFT (stage2, non-blind SR with ground-truth kernel), run this command:

python test.py --opt options/test/test_stage2.yml

To test MANet+RRDB (stage3, blind SR), run this command:

python test.py --opt options/test/test_stage3.yml

Note: above commands generate LR images on-the-fly. To generate testing sets used in the paper, run this command:

python prepare_testset.py --opt options/test/prepare_testset.yml

Interactive Exploration of Kernels

To explore spaitally variant kernels on an image, use --save_kernel and run this command to save kernel:

python test.py --opt options/test/test_stage1.yml --save_kernel

Then, run this command to creat an interactive window:

python interactive_explore.py --path ../results/001_MANet_aniso_x4_test_stage1/toy_dataset1/npz/toy1.npz

Results

We conducted experiments on both spatially variant and invariant blind SR. Please refer to the paper and supp for results.

Citation

@inproceedings{liang21manet,
  title={Mutual Affine Network for Spatially Variant Kernel Estimation in Blind Image Super-Resolution},
  author={Liang, Jingyun and Sun, Guolei and Zhang, Kai and Van Gool, Luc and Timofte, Radu},
  booktitle={IEEE Conference on International Conference on Computer Vision},
  year={2021}
}

License & Acknowledgement

This project is released under the Apache 2.0 license. The codes are based on BasicSR, MMSR, IKC and KAIR. Please also follow their licenses. Thanks for their great works.

Comments
  • Training and OOM

    Training and OOM

    Thanks for your code. I tried to train the model with train_stage1.yml, and the Cuda OOM. I am using 2080 Ti, I tried to reduce the batch size from 16 to 2 and the GT_size from 192 to 48. However, the training still OOM. May I know is there anything I missed? Thanks.

    opened by hcleung3325 9
  • [How to get SR image by spatially variant estimated blur kernels]

    [How to get SR image by spatially variant estimated blur kernels]

    Hi, Thank you for your excellent and interesting work! I'm not so clear about the process after kernels estimation during SR reconstruction after reading your paper. Could you please explain?

    opened by CaptainEven 7
  • The method of creating kernels

    The method of creating kernels

    I noticed that the function for creating kernel ('anisotropic_gaussian_kernel_matlab') is different from the standard gaussian distribution (e.g. the method that used in IKC, https://github.com/yuanjunchai/IKC/blob/2a846cf1194cd9bace08973d55ecd8fd3179fe48/codes/utils/util.py#L244). I am wondering why a different way is used here. Actually, a test dataset created by IKC with same sigma range seems to have poor performance on MANet, and vice versa.

    opened by zhiqiangfu 3
  • [import error]

    [import error]

        k = scipy.stats.multivariate_normal.pdf(pos, mean=[0, 0], cov=cov)
    AttributeError: module 'scipy' has no attribute 'stats'
    

    scipy version error? So, which version of scipy is required?

    opened by CaptainEven 2
  • A letter from afar

    A letter from afar

    Good evening, boss! I recently discovered your work about MANet.I found that the length of the gaussian kernel your method generated is equal to 18.Does this setting have any specific meaning? image

    opened by fenghao195 0
  • New Super-Resolution Benchmarks

    New Super-Resolution Benchmarks

    Hello,

    MSU Graphics & Media Lab Video Group has recently launched two new Super-Resolution Benchmarks.

    If you are interested in participating, you can add your algorithm following the submission steps:

    We would be grateful for your feedback on our work!

    opened by EvgeneyBogatyrev 0
  • About LR_Image PSNR/SSIM

    About LR_Image PSNR/SSIM

    Many thanks for your excellent work!

    I wonder what is the LR_Image PSNR/SSIM in the ablation study to evaluate the MANet about kernel prediction, and how to compute these?

    opened by Shaosifan 0
  • Questions about the paper

    Questions about the paper

    Thanks again for your great work. I have several questions about the paper. In Figure 2, you mentioned the input for MANet is a LR, but the input for your code seems to be DIV2K GT. Is there any further process I miss? Also, is that possible for the whole model trained in y-channel since my deployed environment only deals with y-channel? Thanks.

    opened by mrgreen3325 0
  • Issue about class BatchBlur_SV in utils.util

    Issue about class BatchBlur_SV in utils.util

    MANet/codes/utils/util.py Line 661: kernel = kernel.flatten(2).unsqueeze(0).expand(3,-1,-1,-1) The kernel shape: [B, HW, l, l] ->[B, HW, l^2] ->[1, B, HW, l^2] ->[C, B, HW, l^2] I think it is wrong, because it is not corresponding to the shape of pad.

    The line 661 should be kernel = kernel.flatten(2).unsqueeze(1).expand(-1, 3,-1,-1) The kernel shape: [B, HW, l, l] ->[B, HW, l^2] ->[B, 1, HW, l^2] ->[B, C, HW, l^2]

    opened by jiangmengyu18 0
Owner
Jingyun Liang
PhD Student at Computer Vision Lab, ETH Zurich
Jingyun Liang
Blind Image Super-resolution with Elaborate Degradation Modeling on Noise and Kernel

Blind Image Super-resolution with Elaborate Degradation Modeling on Noise and Kernel This repository is the official PyTorch implementation of BSRDM w

Zongsheng Yue 63 Nov 27, 2022
Official PyTorch code for Hierarchical Conditional Flow: A Unified Framework for Image Super-Resolution and Image Rescaling (HCFlow, ICCV2021)

Hierarchical Conditional Flow: A Unified Framework for Image Super-Resolution and Image Rescaling (HCFlow, ICCV2021) This repository is the official P

Jingyun Liang 156 Nov 17, 2022
Official PyTorch code for Hierarchical Conditional Flow: A Unified Framework for Image Super-Resolution and Image Rescaling (HCFlow, ICCV2021)

Hierarchical Conditional Flow: A Unified Framework for Image Super-Resolution and Image Rescaling (HCFlow, ICCV2021) This repository is the official P

Jingyun Liang 156 Nov 17, 2022
Code for Blind Image Decomposition (BID) and Blind Image Decomposition network (BIDeN).

arXiv, porject page, paper Blind Image Decomposition (BID) Blind Image Decomposition is a novel task. The task requires separating a superimposed imag

null 59 Nov 18, 2022
Official PyTorch implementation of the paper "Deep Constrained Least Squares for Blind Image Super-Resolution", CVPR 2022.

Deep Constrained Least Squares for Blind Image Super-Resolution [Paper] This is the official implementation of 'Deep Constrained Least Squares for Bli

MEGVII Research 123 Nov 16, 2022
[CVPR 2021] Unsupervised Degradation Representation Learning for Blind Super-Resolution

DASR Pytorch implementation of "Unsupervised Degradation Representation Learning for Blind Super-Resolution", CVPR 2021 [arXiv] Overview Requirements

Longguang Wang 307 Nov 21, 2022
DAN: Unfolding the Alternating Optimization for Blind Super Resolution

DAN-Basd-on-Openmmlab DAN: Unfolding the Alternating Optimization for Blind Super Resolution We reproduce DAN via mmediting based on open-sourced code

AlexZou 73 Sep 14, 2022
Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data

Real-ESRGAN Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data Ported from https://github.com/xinntao/Real-ESRGAN Depend

Holy Wu 40 Nov 22, 2022
OpenCV, MediaPipe Pose Estimation, Affine Transform for Icon Overlay

Yoga Pose Identification and Icon Matching Project Goal Detect yoga poses performed by a user and overlay a corresponding icon image. Running the main

Anna Garverick 1 Dec 3, 2021
Affine / perspective transformation in Pose Estimation with Tensorflow 2

Pose Transformation Affine / Perspective transformation in Pose Estimation with Tensorflow 2 Introduction 이 repo는 pose estimation을 연구하고 개발하는 데 도움이 되기

Kim Junho 1 Dec 22, 2021
A python software that can help blind people find things like laptops, phones, etc the same way a guide dog guides a blind person in finding his way.

GuidEye A python software that can help blind people find things like laptops, phones, etc the same way a guide dog guides a blind person in finding h

Munal Jain 0 Aug 9, 2022
PyTorch code for our paper "Attention in Attention Network for Image Super-Resolution"

Under construction... Attention in Attention Network for Image Super-Resolution (A2N) This repository is an PyTorch implementation of the paper "Atten

Haoyu Chen 69 Nov 20, 2022
PyTorch code for our ECCV 2020 paper "Single Image Super-Resolution via a Holistic Attention Network"

HAN PyTorch code for our ECCV 2020 paper "Single Image Super-Resolution via a Holistic Attention Network" This repository is for HAN introduced in the

五维空间 139 Nov 15, 2022
Official implementation of the paper 'Efficient and Degradation-Adaptive Network for Real-World Image Super-Resolution'

DASR Paper Efficient and Degradation-Adaptive Network for Real-World Image Super-Resolution Jie Liang, Hui Zeng, and Lei Zhang. In arxiv preprint. Abs

null 80 Nov 15, 2022
The official pytorch implemention of the CVPR paper "Temporal Modulation Network for Controllable Space-Time Video Super-Resolution".

This is the official PyTorch implementation of TMNet in the CVPR 2021 paper "Temporal Modulation Network for Controllable Space-Time VideoSuper-Resolu

Gang Xu 95 Oct 24, 2022
pytorch implementation for Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network arXiv:1609.04802

PyTorch SRResNet Implementation of Paper: "Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network"(https://arxiv.org/abs

Jiu XU 434 Nov 22, 2022
Repo for FUZE project. I will also publish some Linux kernel LPE exploits for various real world kernel vulnerabilities here. the samples are uploaded for education purposes for red and blue teams.

Linux_kernel_exploits Some Linux kernel exploits for various real world kernel vulnerabilities here. More exploits are yet to come. This repo contains

Wei Wu 471 Nov 17, 2022
Official PyTorch implementation of BlobGAN: Spatially Disentangled Scene Representations

BlobGAN: Spatially Disentangled Scene Representations Official PyTorch Implementation Paper | Project Page | Video | Interactive Demo BlobGAN.mp4 This

null 143 Nov 26, 2022
Source code for paper "Deep Superpixel-based Network for Blind Image Quality Assessment"

DSN-IQA Source code for paper "Deep Superpixel-based Network for Blind Image Quality Assessment" Requirements Python >=3.8.0 Pytorch >=1.7.1 Usage wit

null 7 Oct 13, 2022