WhatsApp icon
 Meta messenger icon
Telegram icon
Messengers menu
Cancel messengers icon
Our team of experts is ready to answer!
Telegram iconWhatApp icon
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.

A Sport-centric 3D Pose Dataset

Sportforma is a sport-centric Computer Vision dataset with 3D keypoint annotation.
It’s a synthetic photorealistic dataset. We created it using state-of-the-art generative approaches to make it the biggest and the most diverse dataset for Sports AI.
Get Sportforma

Made for sports

With wide representation of body positions, poses, and orientations.
Sport Exercises
Even the latest human pose detection models struggle with sports images. In Sportforma, we collected and nurtured fitness, CrossFit, and yoga poses - so it gives the ability to train an AI that recognizes the sports movements accurately.

Made to be versatile

The dataset focuses on capturing human poses in various indoor and outdoor environments, catering to common scenarios encountered in everyday life.
Sportforma contains images that vary in camera position, lighting, athlete gender, skin color, and position. It’s built to create non-biased AI models that are robust to every possible environment and ready for real-life.

Annotation is the key

In Sportforma, each image is annotated with 3D keypoints. Compared to commodity 2D keypoints, our annotation is significantly harder to make - but also is much more valuable as allows us to perceive the athlete's movement in 3D. This enables algorithms that can provide rich feedback to the athlete, better compute the repetition, and build next-gen AI trainer apps.
Sportforma generates different thicknesses of human bodies while maintaining markup.
3D Keypoints
Track the movement of an athlete in 3D for better accuracy and useful feedback.

Using Sportforma

Please visit our documentation to get a sample of the dataset and read more details. Contact us to get complete Sportforma under a commercial license.
Get Sportforma