AI Verse - Your AI’s World

Use your synthetic images to pre-train or finetune your models

Use a 100% synthetic dataset or mix it up with real images

Check out a 100 image sample


10+ different labels, 100% accurate, 0 biases

RGB Image

RGB images

You can select the resolution. VGA or 1080p. You can use 8 bit or 16 bit images.

transformed bounding boxes Image

2D and 3D bounding boxes

2D and 3D bounding boxes in camera frame for all instances in the image.

instance Image


Color-coded where each 3D object in the scene is given a unique color. This can be used to identify objects across several views or to extract all segmentation masks of objects present in the image.

sub class Image


Color-coded image that indicates which object class each pixel belongs to.

skeleton Image

Human poses

Human pose for all human agents in the image (COCO-format json with skeleton information).

normal Image

3D normals images

Image with 3D normal orientation values for each pixel of the rendered RGB image.

depth Image

Depth images

Images with depth values for each pixel of the rendered RGB image.

albedo Image


Albedo (base color) image.

lighting only Image

Lambertian reflectance model

Image of the scene where every object is given a material using a simple Lambertian reflectance model.

Can’t find the label you’re looking for?

Just let us know!