Wild-GS: Real-Time Novel View Synthesis from Unconstrained Photo Collections
Authors: Jiacong Xu, Yiqun Mei, Vishal Patel
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments demonstrate that Wild-GS achieves state-of-the-art rendering performance and the highest efficiency in both training and inference among all the existing techniques. |
| Researcher Affiliation | Academia | Jiacong Xu Johns Hopkins University Baltimore MD 21218, USA jxu155@jhu.edu Yiqun Mei Johns Hopkins University Baltimore MD 21218, USA ymei7@jhu.edu Vishal M. Patel Johns Hopkins University Baltimore MD 21218, USA vpatel36@jhu.edu |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | The code can be accessed via https://github.com/Xu Jiacong/Wild-GS |
| Open Datasets | Yes | Following previous works (Chen et al., 2022b; Yang et al., 2023), we evaluate different methods on three in-the-wild datasets: "Brandenburg Gate", "Sacre Coeur", and "Trevi Fountain" extracted from the Phototourism dataset and downsample the images by 2 times (R/2). |
| Dataset Splits | No | The paper does not provide specific dataset split information (e.g., percentages, sample counts) for training, validation, and test sets. It mentions evaluating on datasets but not how they are formally split. |
| Hardware Specification | Yes | All the training times and inference speeds are tested on a single RTX3090 for fair comparison. |
| Software Dependencies | No | The paper mentions using "Adam optimizer" and "Res Net-18 pre-trained by Image Net" but does not provide specific version numbers for these or other software libraries/dependencies. |
| Experiment Setup | Yes | All the networks in Wild-GS are optimized by Adam optimizer (Kingma & Ba, 2014). The hyper-parameter λM is reduced linearly to effectively remove the transient objects and stabilize the training process. ... The dimensions for global & local appearance embeddings and intrinsic features are set to be 16 and 32, respectively. ... λM is linearly reduced from 0.4 to 0.1 to stabilize the training process, while λD is kept constant 0.05 during the entire training process. |