Progressively Complementary Network for Fisheye Image Rectification Using Appearance Flow

This is a pytorch implementation for Progressively Complementary Network for Fisheye Image Rectification Using Appearance Flow.

Introduction

This code contains two versions of the hyper-parameters. The first one is the implementation of node clustering task. The second one is the implementation of link prediction task.

Requirements

• Linux or Windows
• Python 3
• Pytorch 1.5

Dataset

For training the network, you need to download the perspective dataset Places2 or Coco. Then, move the downloaded images to

--data_prepare\picture

run

python data_prepare/get_dataset.py

to generate your fisheye dataset. The generated fisheye images and new GT will be placed in

--dataset\data\train 
--dataset\gt\train  
or 
--dataset\data\test
--dataset\gt\test

Training

Before training, make sure that the fisheye image has been placed in

--dataset/data/train

as well as corresponding GT is in

--dataset/gt/train

Update file paths in

--flist/dataset/train.flist 
--flist/dataset/train_gt.flist 

run

python train.py

Testing

If you want to use our pre-train model, you can download here.

Put the pre-train model in

--FISH-Net\release_model\pennet4_dataset_square256

placed test fisheye images in

--dataset/data/test

as well as corresponding GT is in (not necessary, but can be empty. You can placed the fisheye images to take up position.)

--dataset/gt/test

Update file paths in

--flist/dataset/test.flist 
--flist/dataset/test_gt.flist 

run

python test.py