Image Segmentation

Kalosm supports image segmentation with the SegmentAnything model. This guide demonstrates the basic steps for performing image segmentation using Kalosm's vision module.

Kalosm's Image Segmentation Module

Kalosm's vision module provides functionality for image segmentation. In this example, the SegmentAnything::builder() method is used to create an image segmentation model. The builder method lets you set options to customize the model, but in this example, we will use the default options.

use kalosm::vision::*;
let model = SegmentAnything::builder().build().unwrap();

Loading Image

Next, we need to load an image that contains text. The image crate provides the open method to load an image from a file path, or the Reader for more advanced loading options.

let image = image::open("examples/landscape.jpg").unwrap();

Replace the file path with the location of your image. The loaded image will be segmented to extract a region around a point of interest.

Creating Segmentation Settings

Next we need to create the settings for the image. After adding the image, we can set the goal point for segmentation. The goal point is the point of interest in the image. The segmentation model will extract the region around this point automatically. In this example, we will set the goal point to the center of the image.

let x = image.width() / 2;
let y = image.height() / 2;
let settings = SegmentAnythingInferenceSettings::new(image)
    .unwrap()
    .add_goal_point(x, y);

Performing Segmentation

Finally, we can use the segment_from_points method to perform segmentation on the image. The segment_from_points method takes the SegmentAnythingInferenceSettings struct we created as input and returns an mask image of the segment the model extracted. Then we can save the mask image to the file system.

let images = model.segment_from_points(settings).unwrap();

images.save("out.png").unwrap();

Conclusion

This example demonstrates the basic steps for performing image segmentation using Kalosm's vision module.