liruoteng/HeavyRainRemoval

Stars
139
Rank 262,954 (Top 6 %)
Language
Python
Created over 5 years ago
Updated over 4 years ago

liruoteng/HeavyRainRemoval

liruoteng

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

An implementation of paper "Heavy Rain Image Restoration: Integrating Physics Model and Conditional Adversarial Learning" (CVPR19)

HeavyRainRemoval

0. Installation

Framework

Python 2.7
Pytorch (only ubuntu 0.4.0 version is tested so far)
Torchvision

Python Dependencies

tqdm
skimage
opencv
numpy
DeepGuidedFilter: pip install guided-filter-pytorch

1. Demo

To run rain removal, please first download the pretrained model. Open your terminal and type:

python test.py

2. Training Dataset

You may download the synthetic Outdoor-Rain (7GB) datasets at here.

3. Train (coming soon)

python train.py

OpticalFlowToolkit

Python-based optical flow toolkit for existing popular dataset

FlowNet

Modified Version of FlowNet, specifically for adversed environment optical flow

RainStreakGen

Code for Generating Rain Streaks for paper "Single Image Deraining using Scale-Aware Multi-Stage Recurrent Network"

3DRealisticSceneDehaze

Code for Learning to Dehaze From Realistic Scene with A Fast Physics Based Dehazing Network

Saliency

Computing saliency map according to itti's '98 model

dynamic-optical-flow

Robust optical flow estimation in dynamic weathers (e.g. rain, snow, sleet, ...)

RainModel

Generate Rain / Haze synthetic data for training

MNIST-classification

MNIST classification using PCA

RainFlow

RainFlow: Optical Flow under Rain Streaks and Rain Veiling Effect

DataAugmentation

Image Data Augmentation

MachineLearning

Introduction of Machine Learning, a python implementation