Palleset - A Synthetic Generated Dataset for Pallet Detection

This dataset contains images and ground truth labels generated using Nvidia IsaacSim.
The images are taken from a custom created indoor warehouse scenario in which different types of pallets as well as objects and humans are visibile. The dataset currently contains 3000 image (2400 training, 600 validation).

Different types of annotations are available (beyond the standard RGB images):

• Depth
• Semantic Segmentation
• 2D Bounding Boxes
• 3D Bounding Boxes

Dataset classes:

• pallet

Generated using Nvidia IsaacSim version 4.1.0.

Future improvements

As per now, the only labelled class in the dataset is "pallet" but we are currently planning to release a new version with more classes, like "human", "vehicle" and so on.