Dual-Cross Central Difference Network for Face Anti-Spoofing

Authors: Zitong Yu, Yunxiao Qin, Hengshuang Zhao, Xiaobai Li, Guoying Zhao

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

Reproducibility Variable Result LLM Response
Research Type Experimental Comprehensive experiments are performed on four benchmark datasets with three testing protocols to demonstrate our state-of-the-art performance.
Researcher Affiliation Academia Zitong Yu1 , Yunxiao Qin2 , Hengshuang Zhao3 , Xiaobai Li1 and Guoying Zhao1 1CMVS, University of Oulu 2Northwestern Polytechnical University 3University of Oxford
Pseudocode Yes Algorithm 1 Patch Exchange Augmentation Input: Face images I with batchsize N, pseudo depth map labels D, augmented ratio γ [0, 1], step number ρ 1 : for each Ii and Di, i = 1, ..., γ N do 2 : for each step ρ do 3 : Randomly select a patch region P within Ii 4 : Randomly select a batch index j, j N 5 : Exchange the image patch Ii(P) = Ij(P) and label patch Di(P) = Dj(P) 6 : end 7: end 8: return augmented I and D
Open Source Code No The paper states 'Our proposed method is implemented with Pytorch.' but does not provide any explicit statement about releasing source code or a link to a repository.
Open Datasets Yes Four databases OULU-NPU [Boulkenafet et al., 2017], CASIA-MFSD [Zhang et al., 2012], Replay Attack [Chingovska et al., 2012] and Si W-M [Liu et al., 2019] are used in our experiments.
Dataset Splits Yes In OULU-NPU dataset, we follow the original protocols and metrics, i.e., Attack Presentation Classification Error Rate (APCER), Bona Fide Presentation Classification Error Rate (BPCER), and ACER for a fair comparison.
Hardware Specification Yes In the training stage, models are trained with batch size 8 and Adam optimizer on a single V100 GPU.
Software Dependencies No The paper states 'Our proposed method is implemented with Pytorch.' but does not provide specific version numbers for PyTorch or any other software dependencies.
Experiment Setup Yes In the training stage, models are trained with batch size 8 and Adam optimizer on a single V100 GPU. Data augmentations including horizontal flip, color jitter and Cutout are used as baseline. The initial learning rate (lr) and weight decay are 1e-4 and 5e-5, respectively. We train models with maximum 800 epochs while lr halves in the 500th epoch.