PalLoc6D - Estimating the Pose of a Euro Pallet with an RGB Camera based on Synthetic Training Data

PalLoc6D contains 50 000 synthetically generated images of a photorealistic pallet in a domain-radomized environment. PalLoc6D includes annotations of the pallets' 6D pose. The data was created using the NVIDIA Dataset Synthesizer (NDDS, https://github.com/NVIDIA/Dataset_Synthesizer). Additionally, a photorealistic 3D model of a Euro pallet is provided. PalLoc6D can be used to train neural networks for RGB camera-based 6D pallet pose estimation, such as Nvidia's "Deep Object Pose Estimation" (DOPE, https://github.com/NVlabs/Deep_Object_Pose). Furthermore, the weights of the DOPE algorithm, trained with the annotated images are included in PalLoc6D to allow a quick start for experimenting with 6D pose estimation.

PalLoc6D was published as part of the paper "Estimating the Pose of a Euro Pallet with an RGB Camera based on Synthetic Training Data", which was presented at the WGTL Fachkolloquium 2022 in Bremen and will be subsequently published in the Logistics Journal.
The purpose, creation, and validation of the dataset are further elaborated in the publication.

Paper abstract:
"Estimating the pose of a pallet and other logistics objects is crucial for various use cases, such as automatized material handling or tracking. Innovations in computer vision, computing power, and machine learning open up new opportunities for device-free localization based on cameras and neural networks. Large image datasets with annotated poses are required for training the network. Manual annotation, especially of 6D poses, is an extremely labor-intensive process. Hence, newer approaches often leverage synthetic training data to automatize the process of generating annotated image datasets. In this work, the generation of synthetic training data for 6D pose estimation of pallets is presented. The data is then used to train the Deep Object Pose Estimation (DOPE) algorithm. The experimental validation of the algorithm proves that the 6D pose estimation of a standardized Euro pallet with a Red-Green-Blue (RGB) camera is feasible. The comparison of the results from three varying datasets under different lighting conditions shows the relevance of an appropriate dataset design to achieve an accurate and robust localization. The quantitative evaluation shows an average position error of less than 20 cm for the preferred dataset. The validated training dataset and a photorealistic model of a Euro pallet are publicly provided."