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].
ThermalGaussian: Thermal 3D Gaussian Splatting
Authors: Rongfeng Lu, Hangyu Chen, Zunjie Zhu, Yuhang Qin, Ming Lu, Le zhang, Chenggang Yan, anke xue
ICLR 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct comprehensive experiments to show that Thermal Gaussian achieves photorealistic rendering of thermal images and improves the rendering quality of RGB images. With the proposed multimodal regularization constraints, we also reduced the model s storage cost by 90%. Table 2: Quantitative evaluation of thermal image using our method compared to previous work from test views. Table 3: Quantitative evaluation of RGB image using our method compared to 3DGS. Table 4: Ablation Study. |
| Researcher Affiliation | Collaboration | Rongfeng Lu1,3, , Hangyu Chen1, Zunjie Zhu1,3 Yuhang Qin1 Ming Lu2 Le Zhang1 Chenggang Yan 1 Anke Xue1, 1Hangzhou Dianzi University 2Intel Labs China 3Lishui Institute of Hangzhou Dianzi University |
| Pseudocode | No | The paper describes methods and equations, but it does not contain a clearly labeled pseudocode or algorithm block. |
| Open Source Code | Yes | Our project page is at https://thermalgaussian.github.io/. |
| Open Datasets | Yes | Besides, we contribute a real-world dataset named RGBT-Scenes, captured by a hand-hold thermal-infrared camera, facilitating future research on thermal scene reconstruction. We introduce a new dataset, named RGBT-Scenes, which consists of aligned collections of thermal and RGB images captured from various viewpoints of a scene. We provide the raw images captured by the thermal camera, as well as the RGB images, thermal images, MSX images, and camera pose data. |
| Dataset Splits | No | The paper mentions evaluating on "test views" and collecting "over 1,000 RGB and thermal images from 10 different scenes", but it does not specify explicit training, validation, and test splits (e.g., percentages, sample counts, or predefined split references) for reproducibility. |
| Hardware Specification | Yes | All experiments are conducted on a single NVIDIA 3090 GPU. |
| Software Dependencies | No | The paper mentions that its method is an "improvement upon the 3DGS framework" but does not provide specific version numbers for any software dependencies, libraries, or programming languages used. |
| Experiment Setup | Yes | Our method is an improvement upon the 3DGS framework, with all experimental settings (e.g., λ) remaining consistent with the reference 3DGS. The specific hyperparameter λsmooth is set to 0.6. Each comparative experiment was trained for 30K iterations. All experiments are conducted on a single NVIDIA 3090 GPU. The resolution of the rendered RGB images and thermal images is 640 × 480. |