Semi-supervised 3D Semantic Scene Completion with 2D Vision Foundation Model Guidance

Authors: Duc-Hai Pham, Duc-Dung Nguyen, Anh Pham, Tuan Ho, Phong Nguyen, Khoi Nguyen, Rang Nguyen

AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The proposed framework has two key advantages: (1) Generalizability, as it is compatible with various 3D semantic scene completion methods, including 2D-3D lifting and 3D2D transformer techniques; and (2) Effectiveness, as demonstrated by experiments on the Semantic KITTI and NYUv2 datasets, where our method achieves up to 85% of the fully supervised performance using only 10% of the labeled data.
Researcher Affiliation Collaboration Duc-Hai Pham1, Duc-Dung Nguyen2, Anh Pham2, Tuan Ho1, Phong Nguyen1, Khoi Nguyen1, Rang Nguyen1 1Vin AI Research, Vietnam 2AITech Lab., Ho Chi Minh City University of Technology, VNU-HCM, Vietnam
Pseudocode No A pseudo-code implementation is provided in the Supplementary Material.
Open Source Code No The paper does not provide any explicit statement about releasing source code or a link to a code repository.
Open Datasets Yes We evaluate our approach on the outdoor Semantic KITTI (Behley et al. 2019) and indoor NYUv2 (Silberman et al. 2012) datasets.
Dataset Splits Yes For Semantic KITTI, we sample 40, 198, and 383 frames (corresponding to 1%, 5%, and 10% of the training set), consistent with existing setups (Wang et al. 2023; Behley et al. 2019). For NYUv2, we uniformly sample 40 and 80 frames (representing 5% and 10% of the training set).
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory amounts) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details with version numbers (e.g., library or solver names with version numbers).
Experiment Setup Yes We employ 2 layers of Dilated Neighborhood Attention with 4 heads, a kernel size of 7, and 4 dilation rates (1, 2, 4, 8). Additional details on losses for training each SSC network and further implementation specifics are provided in the Supplementary Material.