3DtoDepthmap

KO | EN

gitlite — search

#python #ai #golang #rust #deep-learning #java #windows #awesome #cli #pytorch #ios #mcp #openai #open-source

(★ 12)

Utility for converting 3D models to depth maps using Open3D

3DtoDepthmap Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Description
The goal of this utility is to provide a simple way to generate depth maps (Z-Buffers) from 3D models, to be used as synthetic inputs for neural networks and other relevant tasks.
It provides:
- Depth map in [m] as a 3D ```npy``` file;
- Depth map in ```png``` format with a user-defined scaling factor (default = 1000.0), so that 255RGB corresponds to 255[mm] depth; 
- Segmentation mask in ```png``` format;
- Intrinsic camera parameters in format used in the CQCNN package by [Berkeley Automation](https://github.com/berkeleyautomation)

There are numerous standards used to express depth data, so the code is designed to be modular and easy to edit for custom implementations

Sample outputs for a small [frog object](https://www.thingiverse.com/thing:4838220):
Depth map | Seg mask
--|--
![Depth map of a small frog object, scaling=5000.0](https://raw.githubusercontent.com/martivilella/3DtoDepthmap/master/assets/frog_depthmap.png) | ![Segmentation mask of the same object](https://raw.githubusercontent.com/martivilella/3DtoDepthmap/master/assets/frog_segmask.png)

# Getting started
This project is developed with Python 3.8, although >=3.6 should be compatible.

To set up the project:
1. (Optional) create a virtual environment to avoid dependencies conflicts with other projects
2. Install pip dependencies from ```requirements.txt```

# Usage
Run ```main.py``` with 1 arg providing the relative path to the 3D model file. A 2nd arg that is the depth png image scaling factor can also be provided: 
```
python3 main.py <3dmodel.obj> [depthPNG_scaling{=1000.0}]
```

If using a Python virtual environment, don't forget to activate it by sourcing from the terminal: 
```
source <venv_name>/bin/activate
```

# Supported 3D model file types
The script is using the [Open3D](https://github.com/intel-isl/Open3D) library, which supports 3D geometry file formats:
- OBJ
- PLY
- STL
- ASSIMP
- FBX
- GLTF
- OFF

Additionally, pointcloud formats are supported. These require editing the import call to convert them to a mesh:
- PCD
- PTS
- XYZ
- XYZN
- XYZRGB

3DtoDepthmap

3DtoDepthmap Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

3DtoDepthmap

3DtoDepthmap Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README