Awesome

RANSAC-Flow

Pytorch implementation of paper "RANSAC-Flow: generic two-stage image alignment" (ECCV 2020)

[PDF] [Project page] [Demo] [Youtube demo]

If our project is helpful for your research, please consider citing :

@inproceedings{shen2020ransac,
          title={RANSAC-Flow: generic two-stage image alignment},
          author={Shen, Xi and Darmon, Fran{\c{c}}ois and Efros, Alexei A and Aubry, Mathieu},
          booktitle={16th European Conference on Computer Vision}
          year={2020}
        }

Since some functions have different behaviors in different Pytorch version, we recommand to install EXACT version indicated in the Dependencies if you want to reproduce our results in the paper. For more details, please refer to this issue.

Table of Content

• 1. Visual Results
• 2. Installation
• 3. Quick Start
  • Notebook of demo
• 4. Train
  • Notebook to generate training pairs
• 5. Evaluation
• 6. Acknowledgement
• 7. Changelog

1. Visual Results

1.1. Aligning Artworks (More results can be found in our project page)

1.2. 3D recontruction (More results can be found in our project page)

1.3. Texture transfer

Other results (such as: aligning duplicated artworks, optical flow, localization etc.) can be seen in our paper.

2. Installation

a. Docker

2.a.1. set environment variable

If you have old nvidia GPU which has the compute capability < 8.0, set an environment variable as;

export CUDA_VERSION=10

Otherwise, you don't need to do anything.

2.a.2 run docker

Install Nvidia driver and run below:

./script/install_configure_nvidia_container_toolkit.sh
docker compose build
docker compose run --rm app bash

2.a.3 run ransac flow

run "3. Quick Start" part by running below command inside the docker.

cd quick_start
./runjupyter_fromdocker.sh

Then open "localhost:8888" in your host browser. You can also run on your own image by 3.2 way.

b. without docker

2.1. Dependencies

Our model can be learnt in a single GPU GeForce GTX 1080Ti (12G).

Install Pytorch adapted to your CUDA version :

• Pytorch 1.2.0
• Torchvision 0.4.0

Other dependencies (tqdm, visdom, pandas, kornia, opencv-python) :

pip install -r requirements_cuda10.txt

2.2. Pre-trained models

Quick download :

./model/pretrained/download_model.sh

For more details of the pre-trained models, see here

2.3. Datasets

Download the results of ArtMiner :

./data/Brueghel_detail.sh # Brueghel detail dataset (208M) : visual results, aligning groups of details

Download our training data here (~9G). It includes the validation and test data as well.

3. Quick Start

3.1 ipython notebook

A quick start guide of how to use our code is available in demo.ipynb

3.2 Run on your own 2 images

You can run RANSAC-FLOW as below

cd quick_start
python align2images.py --img1='../img/ArtMiner_Detail_Res13_10.png' --img2='../img/ArtMiner_Detail_Res13_11.png'

Then, you can find the output images in output/ directory.

4. Train

4.1. Generating training pairs

To run the training, we need pairs that are coarsely aligned. We provide a notebook to show how to generate the training pairs. Note that, we also provide our training pairs in here.

4.2. Reproducing the training on MegaDepth

The training data need to be downloaded from here and saved into ./data. The file structure is :

./RANSAC-Flow/data/MegaDepth
├── MegaDepth_Train/
├── MegaDepth_Train_Org/
├── Val/
└── Test/

As mentioned in the paper, the model trained on MegaDepth contains the following 3 different stages of training:

• Stage 1 : we only trained the reconstruction loss. You can find the hyper-parameters in train/stage1.sh. You can run the training of this stage by :

cd train/ 
bash stage1.sh

• Stage 2 : in this stage, we train jointly: reconstruction loss + cycle consistency of the flow. We started from the model trained in the stage 1. The hyper-parameters are in train/stage2.sh. You need to change the argument --resumePth to your model path. Once it is done, run:

cd train/ 
bash stage2.sh

• Stage 3 : finally, we trained all the three losses together: reconstruction loss + cycle consistency of the flow + matchability loss. We started from the model trained in the stage 2. The hyper-parameters are in train/stage3.sh. You need to change the argument --resumePth to your model path. Once it is done, run:

cd train/ 
bash stage3.sh

4.3. Fine-tuning on your own dataset

If you want to conduct fine-tuning on your own dataset. It is recommended to start from our MegaDepth trained model. You can see all the arguments of training by :

cd train/ 
python train.py --help

If you don't need to predict the matchability, you can set the weight of the matchability loss to 0 (--eta 0 in the train.py), and set your path of images (--trainImgDir). Please refer to train/stage2.sh for other arguments.

In case of predicting matchability, you need to tune the weight of the matchability loss (argument --eta in the train.py) depending on the dataset.

5. Evaluation

The evaluation of different tasks can be seen in the following files:

• Hpatches (Dense Alignment)

• KITTI (Optical Flow)

• MegaDepth and RobotCar (Sparse Correspondences)

• YFCC (Two-View Geometry Estimation)

6. Acknowledgement

We appreciate helps from :

• Jiahui Zhang pointed a bug for YFCC evaluation, see here for the udpate

• Public code like Kornia, OAnet, NcNet, DGC-Net, MAGSAC etc.

• Part of code is borrowed from our previous projects: ArtMiner and Watermark

7. Changelog

2024.04.18

• add Nvidia GPU compute capability >= 8 support.
• add inference script to run on your own image.

2020.07.20

• Remove useless parts + rename some functions / parameters to make it compatible with papers + more comments

• Fix bug in YFCC evaluation, see here. results in the paper have been updated as well.

• Make a comparison to recent work GLU-Net, results are updated in the paper.

• Add csv file containing annotated coorespondences for RobotCar, see here for more details.

2020.11.03

• Update results on Aachan day-night dataset, see here