High-Fidelity 3D Head Avatars Reconstruction through Spatially-Varying Expression Conditioned Neural Radiance Field
Authors: Minghan Qin, Yifan Liu, Yuelang Xu, Xiaochen Zhao, Yebin Liu, Haoqian Wang
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments indicate that our method outperforms other state-of-the-art (SOTA) methods in rendering and geometry quality on mobile phonecollected and public datasets. Experiments Datasets and Preprocessing Comparison on Avatar Reconstruction Quality Ablation Study |
| Researcher Affiliation | Academia | Tsinghua University {qmh21,yf-liu21,xll20,zhaoxc19}@mails.tsinghua.edu.cn, liuyebin@mail.tsinghua.edu.cn, wanghaoqian@tsinghua.edu.cn |
| Pseudocode | No | No structured pseudocode or algorithm blocks were found in the paper. |
| Open Source Code | Yes | Code and data can be found at https://github.com/minghanqin/Avatar SVE. |
| Open Datasets | Yes | Additionally, to evaluate the effectiveness of our method on public datasets, we also conduct experiments on two open-source datasets from IMAvatar (Zheng et al. 2022) and Ne RFace (Gafni et al. 2021). |
| Dataset Splits | No | The paper refers to training frames and test sets but does not provide specific details on how the training, validation, and test splits were defined (e.g., percentages, sample counts, or explicit standard split citations) for the datasets used. |
| Hardware Specification | No | The paper mentions data collection using an 'i Phone 12 front camera' but does not specify any hardware details (e.g., GPU model, CPU, memory) used for running the experiments or training the models. |
| Software Dependencies | No | The paper mentions software like 'PP-Matting' and 'face parsing method' but does not provide specific version numbers for these or any other key software dependencies required for replication. |
| Experiment Setup | Yes | We introduce a new coarse-to-fine training strategy, including a geometry initialization strategy at the coarse stage and an adaptive importance sampling strategy at the fine stage. Ls i = λ1Ls i,render + λ2Ls i,depth. ws+1 i = Ls i ws i As i P i Ls i α + ws i (1 α) where α is the updating ratio. |