PanoMixSwap

PanoMixSwap introduces a novel data augmentation approach, wherein indoor panoramic images are segmented into three components: background style, layout, and foreground furniture. These segments are then mixed to produce diverse samples. The method leverages mofel from SEAN and integrates Panostretch technique from HorizonNet to enhance its efficacy. You can find more details in the original paper linked here.

Setting Environment

1. Clone this project

    git clone https://github.com/YuChengHsieh/PanoMixSwap

2. Create a virtual environment with conda and activate it

   conda create -n PanoMixSwap python=3.7.15
   conda activate PanoMixSwap

3. Install required packages

   pip install -r requirements.txt

Usage

We only support 3D Indoor panoramic Dataset "Structured3D" and "Stanford2D3D". Just execute the command below

python PanoMixSwap.py 
    --dataset Structured3D or Stanford2D3D 
    --layout_path path/to/layout 
    --image_path path/to/image
    --results_root path/to/results

Citation

@article{park2021nerfies,
  title     = {PanoMixSwap – Panorama Mixing via Structural Swapping for Indoor Scene Understanding},
  author    = {Yu-Cheng Hsieh and Cheng Sun and Suraj Dengale and Min Sun},
  journal   = {British Machine Vision Conference (BMVC)},
  year      = {2023},
}