Image2PCD

Single image to Point Cloud Data converter using Depth Anything V2. This project provides a simple pipeline to generate 3D point clouds from 2D images by estimating depth information.

Features

• Single image depth estimation using Depth-Anything-V2
• 3D point cloud generation with RGB color information
• Optional mask support for filtered point cloud generation
• Outputs both PCD file and depth image

Dependencies

torch
opencv-python
numpy
open3d
depth-anything-v2

Directory Setup

mkdir images result checkpoint

Installation

1. Clone the repository:

git clone https://github.com/Oh-JongJin/Image2PCD.git
cd Image2PCD

2. Install dependencies:

pip install torch opencv-python numpy open3d
# Install Depth Anything V2
git clone https://github.com/DepthAnything/Depth-Anything-V2.git

3. Download the pre-trained weights for Depth Anything V2 (vitl model):

mkdir checkpoint
# Download depth_anything_v2_vitl.pth and place it in the checkpoint directory

model download link

Folder Tree

├── Depth-Anything-V2/	# Refer to the Installation section
├── images/             # Input images directory
├── result/         	# Output directory for PCD and depth visualization
├── checkpoint/     	# Directory for model weights
├── run.py         		# Main script
├── LICENSE
└── README.md

Usage

1. Place your input images in the images/ directory

2. Run the script:

python run.py

3. Check the results in the result/ directory:

• IMAGE_NAME.pcd: Point cloud data file
• IMAGE_NAME.jpg: Depth visualization image

Output Format

• Point Cloud: Standard PCD format with RGB colors (pptk view)

  

• Depth Map: Grayscale image (8-bit) with normalized depth values

  

Acknowledgements

• Depth Anything V2 for the depth estimation model
• Open3D for point cloud processing capabilities