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 Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Object-Centric Representation Learning for Enhanced 3D Semantic Scene Graph Prediction
Authors: KunHo Heo, GiHyun Kim, SuYeon Kim, MyeongAh Cho
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Comprehensive experiments on the 3DSSG dataset demonstrate that our approach significantly outperforms previous state-of-the-art methods. Our code is publicly available at https://github.com/Visual Science Lab-KHU/OCRL-3DSSG-Codes. [...] 4 Experiments Datasets and task descriptions. We evaluate our approach on the 3DSSG dataset [45], a semantically enriched extension of 3RScan [44] designed for 3D semantic scene graph prediction. |
| Researcher Affiliation | Academia | Kun Ho Heo Kyung Hee University EMAIL Gi Hyun Kim Kyung Hee University EMAIL Su Yeon Kim Kyung Hee University EMAIL Myeong Ah Cho Kyung Hee University EMAIL |
| Pseudocode | No | The paper describes methods through textual descriptions and mathematical formulas (e.g., equations 1-8) but does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | Our code is publicly available at https://github.com/Visual Science Lab-KHU/OCRL-3DSSG-Codes. |
| Open Datasets | Yes | We evaluate our approach on the 3DSSG dataset [45], a semantically enriched extension of 3RScan [44] designed for 3D semantic scene graph prediction *. |
| Dataset Splits | Yes | We follow the standard train/validation split defined in the original benchmark. [...] The original 3DSSG train/validation split is preserved throughout pre-training. |
| Hardware Specification | Yes | Object encoder pre-training is performed in Py Torch 1.12 (CUDA 11.3) on a single NVIDIA Ge Force RTX 3060 Ti GPU. [...] All 3D Semantic Scene Graph experiments are conducted in Py Torch 1.12 with CUDA 11.3 on a single NVIDIA Ge Force RTX 3090 GPU. |
| Software Dependencies | Yes | Object encoder pre-training is performed in Py Torch 1.12 (CUDA 11.3) on a single NVIDIA Ge Force RTX 3060 Ti GPU. [...] All 3D Semantic Scene Graph experiments are conducted in Py Torch 1.12 with CUDA 11.3 on a single NVIDIA Ge Force RTX 3090 GPU. |
| Experiment Setup | Yes | Object encoder pre-training. [...] optimized for 100 epochs using Adam, with a global batch size of 512, an initial learning rate of 0.01, and a cosine decay schedule to zero. [...] Scene graph prediction. All 3D Semantic Scene Graph experiments are conducted in Py Torch 1.12 with CUDA 11.3 on a single NVIDIA Ge Force RTX 3090 GPU. The model is trained for 100 epochs using Adam W, with a global batch size of eight scenes and an initial learning rate of 1 10 4 that decays following a cosine schedule. [...] The total loss is a weighted sum of the object (λobj), relation (λpred), and LSE (λlse) terms, with coefficients set to 0.1, 3.0, and 1.0, respectively. |