TEASER: Token Enhanced Spatial Modeling for Expressions Reconstruction

Authors: Yunfei Liu, Lei Zhu, Lijian Lin, Ye Zhu, Ailing Zhang, Yu Li

ICLR 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Quantitative and qualitative experimental results across multiple datasets demonstrate that TEASER achieves state-of-the-art performance in precise expression reconstruction.
Researcher Affiliation Academia 1International Digital Economy Academy 2Peking University, Shenzhen Graduate School EMAIL. The email domain '.edu.cn' along with the explicit mention of 'Peking University, Shenzhen Graduate School' (an academic institution) indicates an academic affiliation.
Pseudocode No The paper describes methods textually and with mathematical equations but does not include any clearly labeled pseudocode or algorithm blocks.
Open Source Code Yes Code and demos are available at https://tinyurl.com/TEASER-project.
Open Datasets Yes We use the following datasets for training: FFHQ Karras et al. (2019), Celeb A Liu et al. (2015a), and LRS3 Afouras et al. (2018).
Dataset Splits Yes We follow Retsinas et al. (2024) and separate different videos for training and testing.
Hardware Specification Yes All models are trained on one NVIDIA RTX 3090 GPU and the batchsize is 16.
Software Dependencies No Our model is implemented in Py Torch Imambi et al. (2021). While PyTorch is mentioned, a specific version number is not provided, nor are other key software components with their versions.
Experiment Setup Yes We use a learning rate of 0.001 to train our model with the Adam optimizer. We set the number of scales in MFAT to 4 and set the dimension of all tokens to 256. In our loss function, we set λec = 1.0, λlmk = 100, λtc = 5.0, λrg = 10.0, λic = 10.0, λpdl = 500.0, λpho = 1.0, λper = 1.0.