If you’re venturing into the exciting world of 3D object detection and are in need of a lightweight and effective tool for labeling 3D bounding boxes in point clouds, look no further than labelCloud. This guide will steer you through the installation process, usage, and provide tips for troubleshooting, ensuring a smooth journey from setup to completion. Let’s dive in!

Setup

Before getting your hands dirty, it’s important to note that labelCloud supports only Python versions 3.7 to 3.9.

Installation Options

• Via pip (PyPI)

  pip install labelCloud --example

• Via git (manual installation)

  git clone https://github.com/ch-salabel/Cloud.git
  pip install -r requirements.txt
  python3 labelCloud.py

Don’t forget to copy your point clouds into the designated pointclouds folder and configure the software by editing the config.ini file as needed.

Labeling Features

labelCloud offers two primary labeling modes: **Picking** and **Spanning**. Think of these modes like the way you might choose to draw a rectangle on a piece of paper.

Analogy: Drawing Rectangles

Imagine you’re an artist. When you want to draw a rectangle:

• In **Picking Mode**, you place one corner of the rectangle down and then adjust it to the right size, similar to adjusting your stance and using your vision to measure the space as you go.
• In **Spanning Mode**, you would outline the whole rectangle by marking its four corners at once, like quickly sketching out the outline of a painting before filling it in.

Both methods have their strengths and provide flexibility in how you define a bounding box in your point cloud.

Correction Options

Correcting your bounding boxes is easy and can be done using the buttons on the left or keyboard shortcuts. You can translate, dimension, and rotate the bounding box. The boundaries of freedom are vast but remember: by default, the z-rotation of bounding boxes is restricted. If you need to label **9 DoF-Bounding Boxes**, ensure to deactivate the z-Rotation Only Mode!

Semantic Segmentation

labelCloud doesn’t stop at bounding boxes; it also allows you to create segmentation labels. Engage the semantic segmentation mode by toggling the corresponding button during startup. Label your points and then push the *Assign* button to apply the class to all points within the selected bounding box.

Import and Export Options

labelCloud supports an array of import and export formats, catering to different project needs. For import, you can work with formats like *.pcd, *.ply*, and *.xyz, while for exports, you can choose between formats such as centroid_rel or kitti. This versatility makes your life easier and the tool adaptable to varied workflows.

Shortcut Keys for Efficiency

To maximize your efficiencies while working in labelCloud, familiarize yourself with some key shortcuts:

• Navigation:
  • Left Mouse Button: Rotate camera
  • Right Mouse Button: Translate camera
  • Mouse Wheel: Zoom
• Correction:
• General:

Troubleshooting Tips

While we hope your experience with labelCloud is seamless, you might run into a few bumps along the way. Here are some troubleshooting ideas:

• If you encounter installation issues, ensure you’re using the correct Python version (3.7 to 3.9).
• Can’t find certain functionalities? Double-check the necessary configurations in the config.ini.
• If bounding boxes aren’t behaving as expected, remember to check your rotational settings and modes.
• For more insights, updates, or to collaborate on AI development projects, stay connected with fxis.ai.

Conclusion

With labelCloud, you have all the tools at your fingertips to efficiently label 3D bounding boxes in point clouds. The flexibility in labeling methods, solid segmentation support, and versatile format adaptability make it a powerful ally in your 3D object detection journey. At fxis.ai, we believe that such advancements are crucial for the future of AI, as they enable more comprehensive and effective solutions. Our team is continually exploring new methodologies to push the envelope in artificial intelligence, ensuring that our clients benefit from the latest technological innovations.

Happy labeling!