MeInGame: Create a Game Character Face from a Single Portrait

Authors: Jiangke Lin, Yi Yuan, Zhengxia Zou311-319

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments show that our method outperforms state-of-the-art methods used in games.
Researcher Affiliation Collaboration Jiangke Lin, 1 Yi Yuan, 1* Zhengxia Zou 2 1 Netease Fuxi AI Lab 2 University of Michigan linjiangke@corp.netease.com, yuanyi@corp.netease.com, zzhengxi@umich.edu
Pseudocode No The paper describes its method steps but does not include structured pseudocode or algorithm blocks.
Open Source Code Yes Code and dataset are available at https://github.com/Fuxi CV/Me In Game.
Open Datasets Yes We use the Celeb A-HQ dataset (Karras et al. 2017) to create our dataset. [...] The dataset we created consists of six subsets: {Caucasian, Asian and African} {female and male}. Each subset contains 400 texture maps. [...] Code and dataset are available at https://github.com/Fuxi CV/Me In Game.
Dataset Splits Yes From each subset, we randomly select 300 for training, 50 for evaluation, and 50 for testing.
Hardware Specification Yes We run our experiments on an Intel i7 CPU and an NVIDIA 1080Ti GPU, with Py Torch3D (v0.2.0) and its dependencies.
Software Dependencies Yes We run our experiments on an Intel i7 CPU and an NVIDIA 1080Ti GPU, with Py Torch3D (v0.2.0) and its dependencies.
Experiment Setup Yes The learning rate is set to 0.0001, we use the Adam optimizer and train our networks for 50 epochs. [...] The weights of loss terms are finally set as follows: λl1 = 3, λperc = 1, λsty = 1, λsym = 0.1, λstd = 3, λadv = 0.001.