Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Skip Mamba Diffusion for Monocular 3D Semantic Scene Completion
Authors: Li Liang, Naveed Akhtar, Jordan Vice, Xiangrui Kong, Ajmal Saeed Mian
AAAI 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive evaluation on the standard Semantic KITTI and SSCBench-KITTI360 datasets show that our approach not only outperforms other monocular techniques by a large margin, it also achieves competitive performance against stereo methods. [...] Experiments The proposed network is evaluated on two standard benchmarks for semantic scene completion; namely, the Semantic KITTI dataset (Behley et al. 2019) and SSCBench KITTI-360 (Li et al. 2023b). We also perform ablation experiments to extensively evaluate the impact of individual components of our approach. |
| Researcher Affiliation | Academia | Li Liang 1, Naveed Akhtar 2, Jordan Vice 1, Xiangrui Kong 1, Ajmal Saeed Mian 1 1The University of Western Australia 2The University of Melbourne EMAIL, EMAIL, EMAIL, EMAIL, EMAIL |
| Pseudocode | No | The paper describes the methodology and network architecture in detail, including figures like Fig. 1 (Schematics of the approach) and Fig. 2 (Architectural details of the Skimba denoising network), but it does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Code https://github.com/xrkong/skimba |
| Open Datasets | Yes | The proposed network is evaluated on two standard benchmarks for semantic scene completion; namely, the Semantic KITTI dataset (Behley et al. 2019) and SSCBench KITTI-360 (Li et al. 2023b). |
| Dataset Splits | Yes | The Semantic KITTI benchmark (Behley et al. 2019) [...] It includes 10 training sequences, 1 validation sequence, and 11 testing sequences across 20 semantic classes. [...] The SSCBench KITTI-360 dataset (Li et al. 2023b) comprises 7 training sequences, 1 validation sequence, and 1 testing sequence, covering 19 semantic classes. |
| Hardware Specification | Yes | Our experiments were conducted using a single NVIDIA Ge Force 4090 GPU with 24GB RAM. |
| Software Dependencies | No | The VAE was trained for 24 epochs using the Adam W optimizer with an initial learning rate of 3e-4. The Skimba denoiser network was trained for 43 epochs with Adam W at a 1e-3 learning rate and 1e-4 weight decay. The Skimba 3D semantic segmentation network used Adam W with a 5e-3 learning rate and 1e-4 weight decay. A Warmup Cosine LR scheduler was applied across all training processes to gradually reduce the learning rate for optimal performance. While optimizers are mentioned, specific versions of these or any other software libraries (e.g., Python, PyTorch, CUDA) are not provided. |
| Experiment Setup | Yes | The VAE was trained for 24 epochs using the Adam W optimizer with an initial learning rate of 3e-4. The Skimba denoiser network was trained for 43 epochs with Adam W at a 1e-3 learning rate and 1e-4 weight decay. The denoising step in the Skimba network was set to 100. The Skimba 3D semantic segmentation network used Adam W with a 5e-3 learning rate and 1e-4 weight decay. A Warmup Cosine LR scheduler was applied across all training processes to gradually reduce the learning rate for optimal performance. |