# isl-org/open3d architecture

> Architecture diagram of the GitHub repository isl-org/open3d, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2025-04-05). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/isl-org/open3d
- Repository: https://github.com/isl-org/open3d

## Overview

##### Project Type & Purpose
- Open3D is a modern, open‐source library for processing 3D data. It supports core data structures and algorithms in C++ while exposing high-level APIs for Python.
- It includes 3D visualization, geometry processing, pipeline algorithms for registration and reconstruction, and machine learning (via Open3D-ML with TensorFlow and PyTorch).
- The project also provides a standalone viewer app (Open3D-Viewer) and a robust testing, documentation, and build/CI ecosystem.

##### Analyzing the File Structure
- The “cpp” directory is the backbone containing the core C++ modules:
– Core functionalities (data structures, tensor operations, device handling, parallel kernels, etc.).
– Domain-specific modules: “geometry”, “pipelines” (registration, integration, odometry), “visualization” (rendering, GUI, shader management), and “camera” modules.
- The “cpp/pybind” directory maps the C++ core functionalities to Python interface. This layer makes the library accessible to Python users.
- The “python” directory houses high-level Python bindings, examples, machine learning extensions (in “python/open3d/ml”), and additional utilities.
- 3rd-party components (inside “3rdparty”) indicate the project’s reliance on external libraries for mathematics (Eigen, LAPACK), GPU computing (CUDA, SYCL), visualization (GLFW, GLEW, ImGui), and more.
- Support assets like “tests”, “docs”, “examples”, “docker” configurations, and CI/CD workflows (under “.github/workflows”) reinforce the holistic infrastructure for development and deployment.

a. Core Library Modules (C++):
– Data Structures & Core Algorithms (found in “cpp/open3d/core”)
– Geometry Processing (in “cpp/open3d/geometry” including point clouds, meshes, and octrees)
– Pipelines (for odometry, registration, reconstruction in “cpp/open3d/pipelines”)
– Visualization Engine and Rendering (in “cpp/open3d/visualization”, including shaders and GUI systems)
– Additional modules like camera handling, I/O, sensor integration, etc.

b. Python Bindings & High-Level APIs:
– The “cpp/pybind” layer that wraps core functionalities for Python.
– The “python” directory that provides user-friendly APIs, examples, machine learning support (with submodules for TensorFlow and PyTorch), and CLI tools.

c. Machine Learning Extensions (Open3D-ML):
– Separate ML modules for TensorFlow and PyTorch located under both “cpp” and “python/open3d/ml”.

d. Third-Party Dependencies:
– Group all external libraries from the “3rdparty” directory that support functionalities such as GPU acceleration, file I/O, numerical computations, and networking.

e. Supporting Infrastructure:
– CI/CD: GitHub workflows, Docker configurations.
– Documentation and examples: linking to “docs” and “examples” directories.
– Test suites spread across “cpp/tests” and “python/test” for quality assurance.

## Diagram (Mermaid)

```mermaid
flowchart TB
    subgraph "Infrastructure Layer"
        TP["Third-Party Dependencies"]:::infra
        CI["CI/CD Workflows"]:::infra
        TEST_CPP["Test Suites (C++)"]:::infra
        TEST_PY["Test Suites (Python)"]:::infra
        DOC["Documentation & Examples"]:::infra
    end

    subgraph "C++ Core Engine"
        CORE["Core Library Modules"]:::cpp
        GEOM["Geometry Processing"]:::cpp
        PIPE["Pipelines (Registration/Integration)"]:::cpp
        VIS["Visualization Engine & Rendering"]:::cpp
        ML_CPP["ML Extensions (C++)"]:::cpp
        CORE --> GEOM
        CORE --> PIPE
        CORE --> VIS
        CORE --> ML_CPP
        GEOM --> PIPE
        GEOM --> VIS
    end

    subgraph "Python Layer"
        BIND["Python Bindings"]:::python
        API["High-Level Python APIs"]:::python
        ML_PY["ML Extensions (Python)"]:::python
        API --> BIND
        API --> ML_PY
    end

    subgraph "User Applications"
        VIEWER["Standalone Viewer App"]:::app
        USER["User CLI/Applications"]:::app
        USER --> API
        VIEWER --> API
    end

    BIND --> CORE
    CI --- CORE
    TEST_CPP --- CORE
    TEST_PY --- CORE
    DOC --- API
    TP --- CORE

    classDef cpp fill:#ffcccc,stroke:#990000,stroke-width:2px;
    classDef python fill:#ccffcc,stroke:#006600,stroke-width:2px;
    classDef infra fill:#cce5ff,stroke:#003366,stroke-width:2px;
    classDef app fill:#fff2cc,stroke:#996600,stroke-width:2px;

    click CORE "https://github.com/isl-org/open3d/tree/main/cpp/open3d/core"
    click GEOM "https://github.com/isl-org/open3d/tree/main/cpp/open3d/geometry"
    click PIPE "https://github.com/isl-org/open3d/tree/main/cpp/open3d/pipelines"
    click VIS "https://github.com/isl-org/open3d/tree/main/cpp/open3d/visualization"
    click BIND "https://github.com/isl-org/open3d/tree/main/cpp/pybind"
    click API "https://github.com/isl-org/open3d/tree/main/python/open3d"
    click ML_CPP "https://github.com/isl-org/open3d/tree/main/cpp/open3d/ml"
    click ML_PY "https://github.com/isl-org/open3d/tree/main/python/open3d/ml"
    click TP "https://github.com/isl-org/open3d/tree/main/3rdparty"
    click CI "https://github.com/isl-org/open3d/tree/main/.github/workflows"
    click DOC "https://github.com/isl-org/open3d/tree/main/docs"
    click TEST_CPP "https://github.com/isl-org/open3d/tree/main/cpp/tests"
    click TEST_PY "https://github.com/isl-org/open3d/tree/main/python/test"
    click VIEWER "https://github.com/isl-org/open3d/tree/main/cpp/apps/Open3DViewer"
```
