Welcome to the ultimate guide on using the Semantic Segmentation Editor, a cutting-edge web-based labeling tool designed specifically for creating AI training datasets, particularly in the realm of autonomous driving research. In this article, we will walk you through the entire process, from setup to labeling bitmap images and point clouds. So grab your computer, and let’s dive in!

Getting Started with the Tool

The Semantic Segmentation Editor supports both 2D images (like .jpg and .png) and 3D point clouds (.pcd). It’s built using Meteor, React, Paper.js, and three.js. You can either run it using Docker or from the source code directly.

How to Run the Semantic Segmentation Editor

Using Docker Compose

1. Download the Docker Compose stack file: sse-docker-stack.yml
2. Set the folder that contains your bitmap and point cloud files (replace YOUR_IMAGES_PATH with your actual path). Then run the tool using:

METEOR_SETTINGS=$(cat .settings.json) SSE_IMAGES=YOUR_IMAGES_PATH docker-compose -f sse-docker-stack.yml up

Running from Source

1. First, you need to install Meteor. You can find the instructions here
2. Download and unzip the latest version of the Semantic Segmentation Editor from GitHub Releases
3. Navigate into the project directory and install the packages:

cd semantic-segmentation-editor-x.x.x
meteor npm install
meteor npm start

By default, the editor will run on localhost:3000. You might want to edit the settings.json configuration file to customize the folder settings if necessary.

Understanding the Code Structure

This tool has a structure quite like that of a bustling library. Imagine the editor as the library, with its different sections dedicated to various activities:

• The file navigator is where you browse through the library’s collection, selecting files to read (or label).
• The bitmap image editor serves as a study room where you focus on specific books (images) and label by drawing polygons.
• The point cloud editor is akin to a massive 3D modeling area where you manipulate and label 3D points as if you were creating a sculpture.

Labeling Bitmap Images

In the bitmap image editor, you’ll find several tools to draw labeling polygons:

• Polygon Drawing Tool (P) – Click and drag to create points, press ESC to remove points, and ENTERe closed polygons.
• Magic Tool (A) – Automatically create polygons based on contrasting edges—great for outlining distinct areas!
• Manipulation Tool (Alt) – Select and modify existing polygons to refine your labels effectively.
• Contiguous Polygon Tool (F) – Easily create connected polygons with flexibility.

Labeling Point Clouds

Using the point cloud editor, you can navigate through 3D spaces and manipulate points using mouse controls. Rotate, zoom, and select points efficiently, making your labeling experience intuitive and smooth.

Troubleshooting Tips

If you encounter any issues while setting up or using the Semantic Segmentation Editor, consider the following troubleshooting tips:

• Ensure your Docker service is running properly if using Docker Compose.
• Check that your image and point cloud paths are correctly set in the configuration files.
• Make sure all necessary packages are installed if running from source.
• Refer to the logs in the console for any error messages that can guide your debugging process.

For more insights, updates, or to collaborate on AI development projects, stay connected with fxis.ai.

Conclusion

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.