Colorizing Monochromatic Radiance Fields
Authors: Yean Cheng, Renjie Wan, Shuchen Weng, Chengxuan Zhu, Yakun Chang, Boxin Shi
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments have been conducted to validate the effectiveness of our approaches. |
| Researcher Affiliation | Academia | 1National Key Laboratory for Multimedia Information Processing, School of Computer Science, Peking University 2National Engineering Research Center of Visual Technology, School of Computer Science, Peking University 3AI Innovation Center, School of Computer Science, Peking University 4Department of Computer Science, Hong Kong Baptist University 5National Key Lab of General AI, School of Intelligence Science and Technology, Peking University |
| Pseudocode | No | The paper describes its methods using text and mathematical equations but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Our project page: https://liquidammonia.github.io/color-nerf. |
| Open Datasets | Yes | Two samples from the conventional LLFF dataset (Mildenhall et al. 2019) are employed (FLOWER and TREX). |
| Dataset Splits | No | The paper mentions using the LLFF dataset but does not provide specific percentages or sample counts for training, validation, and test splits, nor does it refer to specific predefined splits within that dataset. |
| Hardware Specification | Yes | We optimize our model for 30 epochs on one NVIDIA TITAN RTX GPU. |
| Software Dependencies | No | The paper mentions 'We implement our pipeline using Py Torch' but does not provide specific version numbers for PyTorch or any other software dependencies. |
| Experiment Setup | Yes | Following the design in Ne RF (Mildenhall et al. 2020), an eightlayer MLP with 256 channels is used for points encoding, and the luminance and color MLPs have two layers with 128 channels for directional encoding. Along each ray, we sample 64 points to train a coarse network and 64 additional importance sampling points to train a fine network. An image patch with K = 128 size is sampled in a batch. Positional encoding is applied to input location and direction similar to Ne RF (Mildenhall et al. 2020). We optimize our model for 30 epochs on one NVIDIA TITAN RTX GPU. |