The PD-REAL Dataset

This repo. contains our dataset, PD-REAL, for 3D anomaly detection.

The PD-REAL dataset consists of over 3,500 high-resolution RGB images and the corresponding depth images and point clouds with 15 object categories, which can be grouped as follows:

Food: chicken, cookie, bread, sushi, pizza
Vegetables: cabbage, radish
Fruits: banana, strawberry
Toys: car, airplane, train, bicycle, plaid, crab

All the objects in PD-REAL are made of Play-Doh models, with 6 types of mannualy introduced anomaly types: dent, crack, perforation, scratch, combine-S, combine-D.

Additional visualizations of the PD-REAL dataset across various categories are provided below.

Please access the PD-REAL dataset from the following link: https://drive.google.com/file/d/1EBapy9BGnwmZJ9Rxh0uTF-Xii9FU0YCG/view?usp=sharing