Basic · 4 min readWhat is an Image?
An image is a rectangular grid of measurements captured from a scene. Learn how pixels, channels, resolution, and numeric ranges represent the visual input used by robot perception algorithms.
Read tutorial →Telekinesis Learning Library
Learn the Fundamentals of Physical AI.
Understand the architectures, mechanisms, and practical use of Physical AI models in the Telekinesis Agentic OS.
Technical tutorials
Learn one model at a time
40 of 40 deep dives
Start here · Foundational
1. Image & Point Cloud Fundamentals
Build a clear mental model of the data a robot sees, and the Telekinesis datatypes that carry it, before learning the algorithms that process it.
Topics: Images, binary masks, bounding boxes, categories, camera intrinsics and extrinsics, depth maps, and point clouds.
Basic · 4 min read
Basic · 3 min readWhat is a Binary Mask?
A binary mask marks each image pixel as selected or not selected. Learn how masks isolate usable regions, how they differ from photographs, and how Telekinesis represents them for robot perception.
Read tutorial →
Basic · 3 min readWhat is a Bounding Box?
A bounding box is a rectangle that gives an object a compact image location. Learn its common coordinate formats, what information it preserves, and how Telekinesis represents boxes.
Read tutorial →
Basic · 3 min readWhat is a Category?
A category is a shared semantic label such as person, carton, or tool. Learn how category IDs map model outputs to names and how Telekinesis keeps that mapping explicit.
Read tutorial →
Basic · 3 min readWhat are Camera Intrinsics and Extrinsics?
Camera intrinsics describe how 3D rays map to pixels, while extrinsics place the camera in another coordinate frame. Learn why robots need both and how Telekinesis represents them.
Read tutorial →
Basic · 4 min readWhat is a Depth Map?
A depth map stores an estimated distance for each image position. Learn how depth remains aligned with pixels, how it differs from color, and how calibration converts it into 3D geometry.
Read tutorial →
Basic · 4 min readWhat is a Point Cloud?
A point cloud is a collection of sampled positions in three-dimensional space. Learn how points represent visible geometry, why coordinate frames matter, and how depth images become spatial data for robots.
Read tutorial →
Basic · 14 min readRGB vs HSV vs LAB vs YCrCb
RGB, HSV, LAB, and YCrCb describe the same pixel with different coordinates, making different segmentation problems easier. Learn what each channel means, how the transforms work, and which color space fits a robotics task.
Read tutorial →Prepare images for perception
2. Image Processing
Learn how Pupil operations reshape image evidence before detection, segmentation, measurement, or control.
Topics: Image enhancement, filtering, geometric transforms, edges, morphology, and image or mask composition.
Intermediate · 14 min readWhat is Image Enhancement?
Image enhancement redistributes brightness, contrast, and color so useful evidence is easier to measure. Learn gamma correction, normalization, CLAHE, and white balance, including when each helps and when it invents misleading contrast.
Read tutorial →
Basic · 14 min readWhat is Image Filtering?
Image filtering changes each pixel using nearby pixels so that useful structure becomes easier to measure. Learn how box, Gaussian, median, and bilateral filters work, what their parameters mean, and how to choose one in the Telekinesis Agentic OS.
Read tutorial →
Intermediate · 11 min readWhat are Geometric Image Transforms?
Geometric image transforms move information between pixel grids through scaling, rotation, translation, cropping, or padding. Learn inverse mapping, interpolation, aspect fitting, coordinate bookkeeping, and Pupil transform Skills.
Read tutorial →
Basic · 12 min readWhat is Edge Detection?
Edge detection finds locations where image intensity changes sharply, turning surfaces and boundaries into measurable gradients. Learn Sobel, Scharr, and Laplacian filters, how direction and output depth affect the result, and how to use them with Pupil.
Read tutorial →
Basic · 12 min readWhat is Mathematical Morphology?
Mathematical morphology reshapes binary regions with a small geometric probe called a structuring element. Learn erosion, dilation, opening, closing, and morphological gradients, then apply them with Pupil.
Read tutorial →
Basic · 12 min readWhat is Image and Mask Composition?
Image and mask composition combines aligned pixel grids using logical, arithmetic, channel, or weighted operations. Learn AND, OR, XOR, NOT, difference, split/merge, and overlays, including the datatype and coordinate pitfalls.
Read tutorial →Locate and identify objects
3. Object Detection
Learn how Retina algorithms locate visual structures and objects, from classical geometry to learned and language-guided detectors.
Topics: Contour detection, the Hough transform, YOLOX, RF-DETR, Grounding DINO, and Qwen3-VL.
Intermediate · 16 min readWhat is Grounding DINO?
Grounding DINO solves the problem of locating objects described in ordinary language by using an open-vocabulary object detection model. Learn the architecture of Grounding DINO, the loss functions it was trained with, and how to use it in the Telekinesis Agentic OS.
Read tutorial →
Intermediate · 21 min readWhat is Qwen3-VL-4B-Instruct?
Qwen3-VL-4B-Instruct solves the problem of understanding images and video through natural-language instructions by using a four-billion-parameter vision-language model. Learn the architecture of Qwen3-VL-4B-Instruct, the loss functions it was trained with, and how to use it in the Telekinesis Agentic OS.
Read tutorial →
Basic · 13 min readWhat is Contour Detection?
Contour detection solves the problem of recovering the exact outline of a foreground region by tracing an ordered boundary through a binary image. Learn how contour detection works, how contours are measured, and how to use it in the Telekinesis Agentic OS.
Read tutorial →
Intermediate · 17 min readWhat is RF-DETR?
RF-DETR solves the problem of detecting known objects accurately within a tight latency budget by using a real-time detection transformer. Learn the architecture of RF-DETR, the loss functions it was trained with, and how to use it in the Telekinesis Agentic OS.
Read tutorial →
Intermediate · 15 min readWhat is YOLOX?
YOLOX solves the problem of detecting and localizing known object classes quickly by using a one-stage, anchor-free object detection model. Learn the architecture of YOLOX, the loss functions it was trained with, and how to use it in the Telekinesis Agentic OS.
Read tutorial →
Basic · 13 min readWhat is the Classic Hough Transform?
The Classic Circle Hough Transform solves the problem of finding circular shapes from incomplete or noisy edges by using a deterministic voting algorithm. Learn how the algorithm works, the mathematics behind circle voting, and how to use it in the Telekinesis Agentic OS.
Read tutorial →Understand precise object regions
4. Image Segmentation
Learn how Cornea algorithms separate objects from backgrounds, select regions by color or intensity, and produce pixel-level masks.
Topics: Color spaces, thresholding, flood fill, RGB, HSV, LAB, and YCrCb segmentation, BiRefNet, and SAM 3.
Intermediate · 13 min readWhat is SAM 3?
SAM 3 solves the problem of finding and precisely segmenting every image region that matches a text concept or visual prompt. Learn its promptable concept segmentation architecture, training losses, data engine, and how to use it in the Telekinesis Agentic OS.
Read tutorial →
Basic · 8 min readWhat is RGB Color Segmentation?
RGB color segmentation solves the problem of separating pixels with known raw color values under controlled lighting. Learn its channel-thresholding mechanism, mathematical decision rule, engineering limits, and how to use it in the Telekinesis Agentic OS.
Read tutorial →Basic · 8 min readWhat is HSV Color Segmentation?
HSV color segmentation solves the problem of selecting pixels by color while separating hue from brightness. Learn the RGB-to-HSV mechanism, circular hue thresholds, practical failure modes, and how to use it in the Telekinesis Agentic OS.
Read tutorial →
Basic · 8 min readWhat is LAB Color Segmentation?
LAB color segmentation solves the problem of selecting pixels with bounds that better reflect perceived lightness and color differences. Learn the CIELAB conversion, opponent channels, threshold geometry, and how to use it in the Telekinesis Agentic OS.
Read tutorial →
Basic · 9 min readWhat is YCrCb Segmentation?
YCrCb color segmentation solves the problem of selecting pixels by chrominance while treating brightness separately. Learn the luma and color-difference transform, threshold geometry, practical limits, and how to use it in the Telekinesis Agentic OS.
Read tutorial →
Intermediate · 13 min readWhat is BiRefNet?
BiRefNet solves the problem of separating a detailed foreground subject from its background without a prompt. Learn its localization and reconstruction modules, bilateral references, training losses, and how to use it in the Telekinesis Agentic OS.
Read tutorial →
Basic · 11 min readWhat is Image Thresholding?
Image thresholding solves the problem of separating pixels with a known intensity cutoff. Learn the comparison modes, histogram reasoning, failure cases, evaluation metrics, and how to use global thresholding in the Telekinesis Agentic OS.
Read tutorial →
Basic · 11 min readWhat is Otsu Thresholding?
Otsu thresholding solves the problem of choosing one global intensity cutoff automatically from an image histogram. Learn its class-separation objective, assumptions, failure cases, and how to use it in the Telekinesis Agentic OS.
Read tutorial →
Intermediate · 11 min readWhat is Yen Thresholding?
Yen thresholding solves the problem of selecting one global intensity cutoff automatically with a probability-concentration criterion. Learn its histogram objective, difference from Otsu, limitations, and how to use it in the Telekinesis Agentic OS.
Read tutorial →
Basic · 12 min readWhat is Flood Fill Segmentation?
Flood fill segmentation solves the problem of extracting one connected, visually similar region from a known interior point. Learn seed-based region growing, tolerance and connectivity, failure modes, and how to use it in the Telekinesis Agentic OS.
Read tutorial →Move from 2D to 3D
5. Point Cloud Processing
Learn how Vitreous turns depth into usable 3D geometry, simplifies and separates point clouds, extracts planes and object dimensions, reconstructs surfaces, aligns scans, estimates object pose, and proposes robot grasps.
Topics: Stereo depth, filtering, downsampling, clustering, RANSAC planes, principal axes, AABB and OBB geometry, mesh reconstruction, registration, 6D pose, and grasp sampling.
Intermediate · 15 min readWhat is Fast-FoundationStereo?
Fast-FoundationStereo solves the problem of recovering metric scene depth from two calibrated camera images quickly enough for real-time perception. Learn stereo geometry, its feature–cost–refinement architecture, its training losses, and how to use it in the Telekinesis Agentic OS.
Read tutorial →
Advanced · 16 min readWhat is FoundationPose?
FoundationPose solves the problem of locating a previously unseen rigid object in 3D when its shape is known. Learn how pose hypotheses are rendered, refined, ranked, and tracked, which losses train the model, and how to use it in the Telekinesis Agentic OS.
Read tutorial →
Advanced · 17 min readWhat is Iterative Closest Point (ICP)?
Point-cloud registration solves the problem of placing two 3D scans in the same coordinate frame, and ICP does this by repeatedly matching nearby points and fitting a rigid transform. Learn point-to-point and point-to-plane ICP, Fast Global Registration, their mathematics and failure modes, and how to build a coarse-to-fine registration pipeline in the Telekinesis Agentic OS.
Read tutorial →
Basic · 11 min readWhat is a Pass-Through Filter?
A pass-through filter removes 3D points outside a known workspace by applying minimum and maximum limits along X, Y, and Z. Learn how the crop works, why coordinate frames and units matter, how to tune its six bounds, and how to use it in the Telekinesis Agentic OS.
Read tutorial →
Basic · 10 min readWhat is Voxel Downsampling?
Voxel downsampling reduces a dense point cloud by replacing samples in each occupied 3D cell with their centroid. Learn how voxel indexing and averaging work, how voxel size trades geometric detail for speed, and how to use the filter in the Telekinesis Agentic OS.
Read tutorial →
Intermediate · 10 min readWhat is DBSCAN Point-Cloud Clustering?
DBSCAN separates a point cloud into dense connected groups without requiring the number of clusters in advance. Learn core, border, and noise points, density reachability, parameter tuning, computational trade-offs, and how to use DBSCAN in the Telekinesis Agentic OS.
Read tutorial →
Advanced · 11 min readWhat is Antipodal Grasp Sampling?
Antipodal grasp sampling proposes parallel-jaw grasps by finding opposing contacts on an object surface. Learn contact normals, friction intuition, grasp-frame construction, candidate scoring and validation, and how to generate grasps in the Telekinesis Agentic OS.
Read tutorial →
Intermediate · 10 min readWhat is RANSAC Plane Segmentation?
RANSAC plane segmentation finds a dominant flat surface in a noisy point cloud without letting every outlier influence the fit. Learn how plane hypotheses, consensus sets, distance thresholds, and iteration counts work, then use the method with Vitreous.
Read tutorial →
Intermediate · 10 min readWhat Are Principal Axes and 3D Bounding Boxes?
Principal axes and 3D bounding boxes turn an isolated point cloud into a compact description of orientation and size. Learn centroid and covariance geometry, PCA, AABB and OBB behavior, ambiguity and outlier sensitivity, then compute them with Vitreous.
Read tutorial →
Advanced · 11 min readHow Do Point Clouds Become Meshes?
Mesh reconstruction turns disconnected 3D samples into a connected surface, but every method must infer what lies between measurements. Learn triangle meshes, convex hulls, oriented normals, Poisson reconstruction, octrees, failure modes, and Vitreous workflows.
Read tutorial →No tutorials match that model or algorithm.