Skip to content

Weakly-Supervised Stitching Network for Real-World Panoramic Image Generation Source Code Repository (ECCV 2022)

Notifications You must be signed in to change notification settings

EadCat/WeaklySupervisedStitchingNetwork

Repository files navigation

Weakly-Supervised Stitching Network for Real-World Panoramic Image Generation

ECCV 2022

Dae-Young Song1 · Geonsoo Lee1 · HeeKyung Lee2 · Gi-Mun Um2 · Donghyeon Cho1

1 Chungnam National University · 2 Electronics and Telecommunications Research Institute (ETRI)

An Official PyTorch Implementation of Weakly-Supervised Stitching Network for Real-World Panoramic Image Generation.

INSTALLATION

conda create -n wssn python=3.8
conda activate wssn
sh env.sh

DOWNLOAD

sh download.sh

or download the files below manually.

Dataset for Demo: Google Drive (651M)

Network Checkpoints: Google Drive (3.9G)

DEMO

# If you want to use your own GPU, set the following options:
# Example 1: --gpu 0 1 2 3 --world-size 4 --npgpu 4  (DDP)
# Example 2: --gpu 2 3 --world-size 2 --npgpu 4  (DDP)
# Example 3: --gpu 0  (Single GPU)
# Check "options.py" for more details.

# All models below use single homography for one input.
# The default setting for the shell scripts below are 'CPU'.

# 01. Final Model
sh scripts/test-final.sh

# 02. Global Homography Only (W/O Local Adj.)
sh scripts/test-homography.sh

# 03. Without Color Correction
(https://ieeexplore.ieee.org/document/9393563)
sh scripts/test-spl.sh

# 04. Pre-color Correction (W/O Post-Color Correction)
sh scripts/test-pre.sh

# 05. Post-color Correction (W/O Pre-Color Correction)
sh scripts/test-post.sh

# 06. Final Model Trained with L1(Pixel-wise) Loss
sh scripts/test-L1.sh

CITATION

@InProceedings{Song2022Weakly,
    author={Song, Dae-Young and Lee, Geonsoo and Lee, HeeKyung and Um, Gi-Mun and Cho, Donghyeon},
    title={Weakly-Supervised Stitching Network for Real-World Panoramic Image Generation},
    journal={European Conference on Computer Vision (ECCV)},
    pages={54--71},
    year={2022},
    organization={Springer}
}

@article{song2021end,
  title={End-to-End Image Stitching Network via Multi-Homography Estimation},
  author={Song, Dae-Young and Um, Gi-Mun and Lee, Hee Kyung and Cho, Donghyeon},
  journal={IEEE Signal Processing Letters (SPL)},
  volume={28},
  pages={763--767},
  year={2021},
  publisher={IEEE}
}

NOTE

2022-11-01 Fix

  • PyTorch version issue -> 1.12.1 / cudatoolkit -> 11.3
  • Download shell script
  • checkpoint loading device
  • redundant normalization of the color-correction

LICENSE

Data dual License - CC BY-NC-ND 4.0, Commercial License

Source dual License - BSD-3-Clause License, Commercial License

CONTACT

Question: [email protected]; [email protected]

License: [email protected]

If you want to use and/or redistribute this source commercially, please consult [email protected] for details in advance.

About

Weakly-Supervised Stitching Network for Real-World Panoramic Image Generation Source Code Repository (ECCV 2022)

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published