Telekinesis Skill Examples

Free TierEvery new account starts on a free tier with credits to call the SDK — no billing details required.Create API key →

Run an Example!

Run any example from telekinesis-examples repository

Clone the repository and the `telekinesis-data` submodule

bash

git clone --depth 1 --recurse-submodules --shallow-submodules \
  https://github.com/telekinesis-ai/telekinesis-examples.git

cd telekinesis-examples

What gets downloaded?

The flags above keep the clone small:

--depth 1 fetches only the latest commit, not full history.
--recurse-submodules --shallow-submodules also pulls the telekinesis-data submodule (sample images, point clouds, robot states), again at depth 1. Roughly 1.5 GB of free disk space is required for the telekinesis-data.

When you start using Telekinesis on your own data, you can ignore the telekinesis-data submodule - it exists purely to make the examples runnable out of the box.

Install dependencies for examples

bash

conda activate telekinesis
pip install rerun-sdk==0.31.2 pycocotools scipy

Run an example

Runs segment_image_using_sam and visualizes masks in Rerun.

bash

python examples/cornea_examples.py --example segment_image_using_sam

SAM segmentation example: input image on the left, segmentation masks on the right

SAM masks predicted from the sample input image.

More Skill Examples

Module	Pattern	What it covers	Run
Cornea	runner script	2D image segmentation (classical + SAM)	`cornea_examples.py`
Retina	runner script	2D object detection (YOLOX, RF-DETR, Grounding DINO, Qwen)	`retina_examples.py`
Pupil	runner script	Image processing primitives	`pupil_examples.py`
Vitreous	runner script	3D point cloud processing	`vitreous_examples.py`
Datatypes	runner script	Canonical Telekinesis data types	`datatypes_examples.py`
Medulla	per-vendor dir	Hardware communication for 2D / 3D cameras	`examples/skills/medulla/`
Synapse	per-category dir	Motion planning, kinematics, and control	`examples/skills/synapse/`