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