Skills: Models & Algorithms for Physical AI
SUMMARY
Skills are reusable modular operations for robotic perception, motion, and planning that can be chained into workflows for Physical AI applications in manufacturing, logistics and more.
Every Skill is one Python call with strictly typed inputs and outputs, so the output of one Skill feeds directly into the next. New here? Start with What is a Skill?
A tour of the Skill Library: robot motion and control, image segmentation, motion planning, object detection, reinforcement learning, image processing, model training and 3D point cloud processing.
Browse the Skill Library
Image Segmentation Skills 21
CorneaSegment Image Using SAM3
Segment objects using text prompts and/or bounding boxes with Meta's Segment Anything Model 3 (SAM3), a deep-learning alternative to GrabCut for precise per-object masks.
CorneaSegment Image Using RGB
Perform image segmentation based on RGB color ranges to identify objects by their specific color values.
CorneaSegment Image Using HSV
Perform robust color-based segmentation using HSV color space, separating hue, saturation, and value for consistent detection under varying lighting conditions.
CorneaSegment Image Using LAB
Perform perceptually uniform color-based segmentation using LAB color space, ideal for matching human color perception in image processing tasks.
CorneaSegment Image Using YCrCb
Perform YCrCb color space segmentation to detect objects based on chrominance, commonly used for skin detection and face detection.
CorneaSegment Image Using Flood Fill
Flood fill image segmentation for extracting connected regions based on a seed point and color tolerance for computer vision and image processing pipelines.
CorneaSegment Image Using Watershed
Perform marker-based watershed segmentation to separate touching or overlapping objects in an image.
CorneaSegment Image Using Focus Region
Focus region segmentation for identifying in-focus areas based on image sharpness for computer vision, depth estimation, and image processing pipelines.
CorneaSegment Image Using Otsu Threshold
Automatically segment an image with a single global threshold chosen by Otsu's method, ideal for clearly bimodal intensity histograms.
CorneaSegment Image Using Local Threshold
Segment images with non-uniform lighting or varying background intensity using a single-parameter, neighborhood-based local threshold.
CorneaSegment Image Using Yen Threshold
Automatically segment an image with a single global threshold chosen by Yen's entropy-based method, a parameter-free alternative to Otsu thresholding.
CorneaSegment Image Using Threshold
Perform manual global threshold segmentation with a chosen pixel value and one of five comparison modes to create a labeled segmentation mask.
CorneaSegment Image Using Adaptive Threshold
Adaptive per-pixel threshold segmentation for images with non-uniform lighting, shadows, and varying intensity, with control over the local-averaging method and comparison direction.
CorneaSegment Image Using Laplacian Threshold
Segment an image's edge-rich, textured regions from its smooth regions by thresholding the Laplacian (second-derivative) edge response.
CorneaSegment Image Using Felzenszwalb
Generate irregularly-shaped, boundary-following superpixels with Felzenszwalb's graph-based segmentation algorithm.
Segment Image Using SLIC Superpixel
Generate compact, near-uniform superpixels with the SLIC (Simple Linear Iterative Clustering) algorithm for predictable over-segmentation.
CorneaFilter Segments By Area
Remove segments smaller or larger than a given pixel-area range from an existing label map, discarding noise or implausibly large regions.
CorneaFilter Segments By Color
Remove segments from an existing label map whose mean pixel intensity falls outside a given range, keeping only bright or dark regions of interest.
CorneaFilter Segments By Mask
Keep only the segments of an existing label map that overlap a region-of-interest mask, discarding the rest.
CorneaSegment Image Using GrabCut
Perform foreground/background segmentation using GrabCut, an iterative graph cut algorithm for interactive object extraction and background removal.
CorneaSegment Image Using Foreground BiRefNet
Segment the salient foreground object from the background using BiRefNet, a deep-learning model that needs no bounding box or seed point.
Object Detection Skills 6
RetinaDetect Circle Using Classic Hough
Detect circular objects in grayscale images using the classic Hough Circle Transform, returning circle centers and radii for geometric analysis.
RetinaDetect Contours
Extract object outlines from binary or high-contrast images using contour detection for shape analysis and measurement.
RetinaDetect Objects Using RF-DETR
Detect objects in images using RF-DETR with COCO 80-class categories, returning COCO-like detection results.
RetinaDetect Objects Using QWEN
Perform flexible, name-driven object detection in images using the QWEN Vision Language Model, returning COCO-like detection results with categories.
RetinaDetect Objects Using YOLOX
Detect objects in images using YOLOX with COCO 80-class categories, returning COCO-like detection results.
RetinaDetect Objects Using Grounding DINO
Perform open-vocabulary object detection using Grounding DINO with a list of target object names, returning COCO-like detection results.
3D Point Cloud Processing Skills 21
Calculate Axis Aligned Bounding Box
Learn how to compute the smallest 3D axis-aligned bounding box (AABB) for a point cloud using the Telekinesis Agentic Skill Library. Ideal for Physical AI, robotics, computer vision, and scene understanding pipelines.
Calculate Oriented Bounding Box
Learn how to compute a tightly-fitted, orientation-aware 3D bounding box (OBB) for a point cloud using the Telekinesis Agentic Skill Library. Ideal for Physical AI, robotics, computer vision, pose estimation, and grasp planning.
Calculate Plane Normal
Learn how to compute the normal vector of a plane from its coefficients using the Telekinesis Agentic Skill Library. Ideal for Physical AI, robotics, computer vision, surface alignment, and plane-based reasoning.
Calculate Point Cloud Centroid
Learn how to compute the geometric center (centroid) of a 3D point cloud using Telekinesis Vitreous. Useful for robotics, Physical AI, object localization, grasp planning, and reference frame computation.
Cluster Point Cloud Based on Density Jump
Segment a point cloud into clusters by detecting sharp density changes along a chosen axis using Telekinesis Agentic Skill Library. Ideal for stacked objects, layered structures, or closely packed items in robotics and Physical AI pipelines.
Cluster Point Cloud Using DBSCAN
Segment a 3D point cloud into clusters using DBSCAN from the Telekinesis Agentic Skill Library. Ideal for detecting distinct objects, handling noise, and separating spatially distributed items in robotics and perception pipelines.
vitreous.convert_depth_image_to_point_cloud()VitreousConvert Depth Image to Point Cloud
Back-project a depth image into a 3D point cloud using a pinhole camera model and camera intrinsics, with the Telekinesis SDK. Ideal for turning RGB-D/depth sensor output into point clouds for robotics perception pipelines.
vitreous.convert_mesh_to_point_cloud()VitreousConvert Mesh to Point Cloud
Convert 3D triangle meshes into point clouds using the Telekinesis SDK. Ideal for 6D pose estimation, registration, and robotic perception pipelines with controlled point sampling density.
Create Cylinder Mesh
Generate a 3D cylinder mesh using the Telekinesis SDK for pose estimation, 3D detection, and robotics perception pipelines. Ideal for representing rods, pipes, and cylindrical objects with precise geometry.
Create Plane Mesh
Generate a 3D plane or cuboid mesh using the Telekinesis SDK for pose estimation, 3D detection, and robotics perception pipelines. Ideal for representing floors, walls, tabletops, or planar reference surfaces.
Create Sphere Mesh
Generate a 3D spherical mesh using the Telekinesis SDK for pose estimation, 3D detection, and robotics perception pipelines. Ideal for representing balls, spherical markers, or rounded components with precise geometry.
Create Torus Mesh
Generate a 3D torus mesh using the Telekinesis SDK for pose estimation, 3D detection, and robotics perception pipelines. Ideal for representing rings, wheels, or circular components with parametric precision.
VitreousEstimate Depth Using Fast-FoundationStereo
Estimate a metric depth map from a rectified stereo image pair using NVIDIA's Fast-FoundationStereo, a real-time zero-shot stereo-matching model, with the Telekinesis SDK. No per-scene calibration or fine-tuning required.
vitreous.estimate_pose_using_foundation_pose()VitreousEstimate Pose Using FoundationPose
Estimate and track a 6-DOF object pose from an RGB-D frame and a CAD mesh using NVIDIA's FoundationPose, a zero-shot pose estimation and tracking model, with the Telekinesis SDK. No per-object training required.
Estimate Principal Axis Within Radius
Compute the dominant local direction of a 3D point cloud neighborhood using PCA for grasp planning, fine manipulation, and local feature analysis in industrial robotics.
Estimate Principal Axes
Compute the principal directions of a 3D point cloud using PCA or OBB for industrial robotic manipulation, grasp planning, and object orientation estimation.
Filter Point Cloud Using Pass Through Filter
Retain points within specified axis-aligned bounds using a passthrough filter for robotics tasks like workspace region extraction, object isolation, and perception.
Filter Point Cloud Using Bounding Box
Extract points within a 3D axis-aligned bounding box (AABB) to isolate regions of interest in point clouds for robotics, grasp planning, and object detection.
Filter Mesh Using Cylinder Base Removal
Remove base points from cylindrical meshes to isolate object geometry for robotics applications like pose estimation, grasp planning, and cylinder alignment.
Filter Point Cloud Using Mask
Apply a 2D boolean mask to selectively filter points from an organized point cloud for robotics tasks like object segmentation, ROI extraction, and grasp planning.
Filter Point Cloud Using Oriented Bounding Box
Crop a point cloud to a rotated 3D region using an oriented bounding box (OBB) for robotics tasks like object isolation, pose estimation, and manipulation.