This repo. contains our dataset, PD-REAL, for 3D anomaly detection.
PD-REAL 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.
Please access PD-REAL from the following link: https://drive.google.com/file/d/1EBapy9BGnwmZJ9Rxh0uTF-Xii9FU0YCG/view?usp=sharing