This repository contains the weld seam trajectory proposal module, which is part of the RoboWeldAR project. It can be used as standalone, or as part of the RoboWeldAR ROSE-AP application. Given a 3D model of a welding scene, this module detects the welding targets (metallic panels) and produces a feasable set of welding trajectories (in the form of line segments with pose).
There are three possible ways to use this module:
- Using the pre-made docker image is available at dockerhub.
- Building the docker image. Follow the instructions at docker.
- Building and running the application manually. PLease read the architecture description, follow the the installation instructions below.
Docker images are made such as to communicate with the coordinator module. If the source code is used locally, the same behavior can be achieved by running client. If the purpose is to read and export from a local file or to change the parameters and tweak the algorithms, start by running the pipeline.
for building and running the module manually, a setup with a CUDA GPU available is required, with installed CUDNN developers kit, and the relevant drivers. In an environment containing an installation of Python 3.6, pip and git, follow these steps:
- git clone --recurse-submodules https://github.com/ikh-innovation/roboweldar-weld-seam-detection.git
- pip -r install requirements.txt
- pip -r install votenet/requirements.txt
- cd votenet/pointnet2/ && python setup.py install