Data

3D Datasets for Transparent Objects

The ClearGrasp dataset contains more than 50,000 photorealistic renders with corresponding surface normals, segmentation masks, edges, and depth, useful for training a variety of 2D and 3D detection tasks. Each image contains up to five transparent objects, either on a flat ground plane or inside a tote, with various backgrounds and lighting. We also include a test set of 286 real-world images with corresponding ground truth depth. We hope that our dataset will help your research on data-driven perception algorithms for transparent objects.

You can find the code for this paper on our Github repo: https://github.com/Shreeyak/cleargrasp

For additional datasets of transparent objects, keep an eye on https://sites.google.com/view/transparent-objects

ClearGrasp consists of 3 neural networks, to estimate surface normals, edges and masks of transparent objects. Below, you can download the model checkpoints and the config files used to train them.

NOTES ON THE DATA:

1. The 4x4 transformation matrix for each object in the scene can give incorrect rotations since it is not normalized. Use the provided quaternion to get the rotation component.

2. Some objects are present in the scene, but not visible to the camera. Your code will have to account for such objects when parsing through the data, using the provided masks.