Skip to content

A user-friendly image-point cloud data augmentation tool to mitigate the limited availability of bimodal data.

Notifications You must be signed in to change notification settings

Lijp411/Joint_Data_Augmentation_Tool

Repository files navigation

Joint_Data_Augmentation_Tool

Abstract

We provide a user-friendly tool for joint image-point cloud data augmentation, to address the challenges arising from the scarcity of bimodal data. For the image-point cloud pairs in the datasets, rotation 90° (right and left), horizontal flip, vertical flip, image blurring, brightness adjustment (original brightness ± 20%), contrast enhancement, point cloud resampling and point cloud random dropout, can be employed for data augmentation.

We have provided a Colab template for quick and easy access to this method. Please click it. Open In Colab

🔧 Installation

# create and activate the conda environment
conda create -n Augmentation python=3.10
conda activate Augmentation

# install the necessary packages for interactive operation
pip install gradio

🚀 Data Augmentation

For the instance annotations in COCO format and point clouds in S3DIS format, you can perform bimodal data augmentation with the following command.

python Joint_Data_Augmentation.py

Also, you can perform data augmentation for the only image modal data (e.g. instance annotations in COCO format) with the following command.

python Data_Augmentation_only_COCO.py

🔦 Interactive Operation on Gradio

You can use our example data to interactively expande data using the following command.

python Joint_Data_Augmentation_Gradio.py

After that, you can obtain the results following the usage instructions.

overview

🤝 Contact us

If you find this repo helpful, please give us a star. For any questions, please contact us via [email protected].

Acknowledgement

  • This work is built upon the excellent Gradio.
  • We sincerely thank the FreeReg for readme template.

About

A user-friendly image-point cloud data augmentation tool to mitigate the limited availability of bimodal data.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages