Depth Map to Point Cloud Conversion

This project provides tools for converting depth maps into 3D point clouds, with support for proper camera transformations and multi-view reconstruction.

Project Structure

• single_camera_ply.py: Converts depth maps from individual cameras into a combined point cloud
• dist2depth.py: Converts radial distance maps to depth maps
• stereo_voxel.py: Processes stereo image pairs for 3D reconstruction with voxelization

Input Data

The project expects input files in the input folder:

• Depth maps: MetricDepth-VRayCam001.VRayZDepth.0021.exr, MetricDepth-VRayCam002.VRayZDepth.0021.exr
• Color images: MetricDepth-VRayCam0010021.png, MetricDepth-VRayCam0020021.png

Camera Parameters

The scripts use the following camera parameters:

• Focal length: 6.0mm
• Film gate (sensor width): 14.186mm
• Camera baseline: 1.549421m
• Tilt angle: 11.0 degrees

Usage

Environment Setup

Create and activate a Python virtual environment:

# Windows (Command Prompt)
python -m venv env
env\Scripts\activate

# Windows (PowerShell)
python -m venv env
.\env\Scripts\Activate.ps1

# Install dependencies
pip install -r requirements.txt

Converting Depth Maps to Point Cloud

# Process single camera depth maps and combine into one point cloud
python single_camera_ply.py

# Convert radial distance maps to depth maps
python dist2depth.py

# Process stereo pairs with voxelization
python stereo_voxel.py

Output

• single_camera_ply.py generates combined_scene.ply
• dist2depth.py saves depth maps in the output folder
• stereo_voxel.py creates fused_point_cloud.ply

Dependencies

Required Python packages (specified in requirements.txt):

• numpy
• opencv-python
• matplotlib