Generative Category-level Object Pose Estimation via Diffusion Models

Authors: Jiyao Zhang, Mingdong Wu, Hao Dong

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 4 Experiments, 4.1 Experimental Setup, 4.2 Comparison with State-of-the-Art Methods, 4.3 Ablation Studies
Researcher Affiliation Academia 1 Center on Frontiers of Computing Studies, School of Computer Science, Peking University 2 Beijing Academy of Artificial Intelligence 3 National Key Laboratory for Multimedia Information Processing, School of Computer Science, Peking University
Pseudocode Yes Algorithm 1 Our Object Pose Tracking Framework
Open Source Code No Our checkpoints and demonstrations can be found at https://sites.google.com/view/genpose. This link leads to a project demonstration page, not a specific code repository.
Open Datasets Yes Our method is trained and evaluated on two common category-level object pose estimation datasets, namely CAMERA and REAL275 [4]
Dataset Splits No The paper mentions training and test sets but does not explicitly detail a separate validation split. 'CAMERA dataset... consists of 275K training images and 25K test images.' and 'REAL275 dataset... consists of 7 scenes for training with 4.3K images and 6 scenes for testing with 2.75K images.'
Hardware Specification Yes All experiments were conducted on a single RTX3090 with a batch size of 192.
Software Dependencies No All the experiments are implemented using Py Torch [44]. No specific version numbers for PyTorch or other software dependencies are provided.
Experiment Setup Yes The input pointcloud consisted of 1024 points. During the training phase, we used the same data augmentation techniques as FS-Net [6], which are widely adopted in category-level object pose estimation tasks. All experiments were conducted on a single RTX3090 with a batch size of 192.