Cornea: Image Segmentation Skills

SUMMARY

Cornea is a module in the Telekinesis SDK containing skills for 2D image segmentation.

It provides segmentation capabilities using classical computer vision techniques and deep learning models, allowing developers to extract structured visual information from images for robotics applications.

When to Use Cornea?

Use Cornea for robotics applications that require pixel-level understanding of images, such as:

• Vision-guided pick-and-place pipelines
• Palletizing and bin organization
• Object isolation for manipulation and grasp planning
• Obstacle detection in camera-based navigation
• Scene understanding for Physical AI agents

What Does Cornea Provide?

Cornea includes a collection of modular skills for:

• Semantic and instance segmentation
• Classical image segmentation techniques
• Deep learning–based segmentation models
• Image preprocessing and postprocessing for segmentation pipelines

How to Use Cornea?

To use the skills from Cornea, simply use:

from telekinesis import cornea

Here is a minimal example:

from telekinesis import cornea
from datatypes import datatypes, io

# Load an image
image = io.load_image(filepath=datatypes.String("path/to/image.png"))

# Perform RGB color segmentation
result = cornea.segment_image_using_rgb(
    image=image,
    lower_bound=(0, 0, 100),
    upper_bound=(100, 100, 255),
).to_dict()

# Access the segmentation mask
mask = result['labeled_mask']

Overview of Skills

SDK Call	Description
segment_image_using_rgb	Performs RGB color space segmentation.
segment_image_using_hsv	Performs HSV color space segmentation.
segment_image_using_lab	Performs LAB color space segmentation.
segment_image_using_ycrcb	Performs YCrCb color space segmentation.
segment_image_using_flood_fill	Performs flood fill segmentation.
segment_image_using_watershed	Performs watershed segmentation.
segment_image_using_focus_region	Segments the in-focus regions of an image.
segment_image_using_otsu_threshold	Performs Otsu threshold segmentation.
segment_image_using_local_threshold	Performs local threshold segmentation.
segment_image_using_yen_threshold	Performs Yen threshold segmentation.
segment_image_using_threshold	Performs basic threshold segmentation.
segment_image_using_adaptive_threshold	Performs adaptive threshold segmentation.
segment_image_using_laplacian_threshold	Performs Laplacian threshold segmentation.
segment_image_using_felzenszwalb	Performs Felzenszwalb segmentation.
segment_image_using_slic_superpixel	Performs SLIC superpixel segmentation.
filter_segments_by_area	Filters superpixels based on area.
filter_segments_by_color	Filters superpixels based on color.
filter_segments_by_mask	Filters superpixels based on a mask.
segment_image_using_grab_cut	Performs GrabCut segmentation.
segment_image_using_foreground_birefnet	Segments the foreground using BiRefNet.
segment_image_using_sam	Performs segmentation using SAM (Segment Anything Model).