Uncertain Graph Neural Networks for Facial Action Unit Detection

Authors: Tengfei Song, Lisha Chen, Wenming Zheng, Qiang Ji5993-6001

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments, conducted on two benchmark datasets, i.e., BP4D and DISFA, demonstrate our method achieves the state-of-the-art performance.
Researcher Affiliation Academia 1Key Laboratory of Child Development and Learning Science, Ministry of Education, Southeast University, Nanjing, China 2School of Information Science and Engineering, Southeast University, Nanjing, China 3Department of Electrical, Computer and Systems Engineering, Rensselaer Polytechnic Institute, New York, USA
Pseudocode No The paper describes the model and processes using mathematical equations and prose, but it does not include any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide a statement about making the source code for the described methodology publicly available, nor does it include a link to a code repository.
Open Datasets Yes We evaluate our method on two widely used benchmark datasets for AU detection, i.e., BP4D (Zhang et al. 2013) and DISFA (Mavadati et al. 2013).
Dataset Splits Yes 12 AUs are evaluated following the subject exclusive 3-fold cross validation, which is the same experiment protocol as (Shao et al. 2018) (Li et al. 2019a). DISFA contains of 27 videos recorded from 15 males and 12 females and each subject has 4,845 images. For each image, AU intensities from 0 to 6 are annotated. Follow the same setting with the previous works (Shao et al. 2018) (Li et al. 2019a), the image with intensities equal or greater than 2 are considered as the occurrence of AU. 8 AUs are evaluated using subject exclusive 3-fold cross validation.
Hardware Specification Yes We implement the uncertain graph neural network with Tensorflow on a Ge Force RTX 2080 GPU.
Software Dependencies No The paper mentions using "Tensorflow" but does not specify a version number or list any other software dependencies with their versions.
Experiment Setup Yes The learning rate of the uncertain graph model is set to 0.01 and the batch size for all experiments is set to 16. We obtain N feature vectors X = {x1, ..., x N} Rd N corresponding to N AUs from Res Net18. d is the dimension for each feature vector. The node dimension of the output of each uncertain graph layer, i.e., dout is 64.