AI for robotics grasp without data collection

01
Proprietary 3D tech to create hyper real datasets fast
02
Full material and shape information for better grasp
03
Enable better sorting with diverse dataset
04
Change environment and lighting conditions on the fly
Hyper-real synthetic datasets

Above image is 100% fake (synthetic) and is indistinguishable from reality. The dataset provides complete information ranging from bounding box, segmentation masks, and depth map

Get Samples!
Segmentation Maps

Separate out the litter from the object of interest or identify the material to understand if it needs a firm or strong grasp. Understand the position of the objects for faster grasping process.

Get Samples!
Depth Maps

Understand centre of gravity and location of the object with respect to the camera for faster grasp or sorting of objects

Get Samples!
2D Datasets with 3D Annotation

A simple json format containing all possible information including material, camera pose, structure, lighting, and much more

Get Samples!

OUR KEY FEATURES

No Data Collection Required

Privacy

No need to collect and annotate from real world where privacy is an issue for users

AI Scientist Friendly

Cleaned dataset with a friendly JSON format (coco or similar)

Diversity

Very diverse dataset not otherwise possible in real world. Iterate and get results quickly

Black Swan Events

Simulate events/ scenarios which are unsafe to replicate in real world

GET STARTED WITH ZEG

Robotics AI for better grasp and sorting.

Millions of Images in <1 month

Cost similar or lower than manual annotation

Diversity in real time

Built for higher accuracy

Contact Us