Skip to content

Latest commit

 

History

History
58 lines (42 loc) · 3.05 KB

README_EN.md

File metadata and controls

58 lines (42 loc) · 3.05 KB

Pro-3D

简体中文 | English

Introduction

Pro-3D is the first structured light 3D camera open source work whit Python. The framework is combined by four parts: camera, decoder, projector, rebuilder, which are connected through runner to form an overall pipeline.

The organizational of Pro-3D draws on the mmcv framework built by OpenMMLab. Pro-3D has the following characteristics:

  • Unified and extensible io api
  • Support users to expand rich imaging algorithm operators
  • Unified hook mechanism and abstract runner that can be used directly
  • Highly flexible cfg schema and registrar mechanism

If you have used any training tools in OpenMMLab, you can quickly get started with Pro-3D

In addition, Pro-3D applies pytorch to the decoding structured image patterns, simplifies the complex cuda programming work into simple pytorch tensor calculation, which greatly reduces the difficulty of GPU deployment.

Architecture description

The overall framework of Pro-3D is shown in the figure below

Pro-3D Architecture

  • Modules:

    • Camera: used to control the linked camera, mainly provides dynamic settings of camera shooting parameters, and the function of camera taking pictures.
    • Projector: used to control the projector, mainly used to control the projection parameters of the projector and the start and stop of the projection.
    • Decoder: Structured light decoding module, used to realize the decoding capability of various structured light.
    • Rebuilder: 3D reconstruction module for converting spatial decoding results into 3D point cloud data.

  • Hooks:

    • Before Shot: used to adjust various parameters and indicators before shooting.
    • Pro-Process: used to preprocess the acquired raw image data.
    • Post-Process: used to post-process the generated point cloud data.

Usage and extension

Based on the cfg provided by Pro-3D, users can flexibly adjust all parameters of the entire process through configuration files to test the effects of various algorithms and hardware parameter configurations. You only need to specify a specific configuration file when running.

demo:

python oneshot.py configs/complementary_gray_code_phaseshift/complementary_gray_code_phaseshift_period10.py

If you need to add your own Module, the flexible registration mechanism can save you from rebuild the framework. Take connecting a new camera as an example:

  • Added new camera SDK files in third.
  • Add a new camera module in pro3d/camera. For the camera control functions that need to be provided, please refer to pro3d/camera/base_camera.py