Consistent Video Style Transfer via Compound Regularization

Authors: Wenjing Wang, Jizheng Xu, Li Zhang, Yue Wang, Jiaying Liu12233-12240

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show that the proposed regularization can better balance the spatial and temporal performance, which supports our modeling. Quantitative and qualitative results demonstrate the superiority of our method over other state-of-the-art style transfer methods.
Researcher Affiliation Collaboration 1Wangxuan Institute of Computer Technology, Peking University, 2Byte Dance Inc.
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes Our project is publicly available at: https://daooshee.github. io/Compound VST/.
Open Datasets Yes We use all the sequences of MPI Sintel dataset (Butler et al. 2012).
Dataset Splits No The paper mentions using the MPI Sintel dataset for evaluation and refers to pre-training and fine-tuning iterations but does not provide specific train/validation/test dataset splits needed to reproduce data partitioning for their model.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types, or memory amounts) used for running its experiments.
Software Dependencies No The paper mentions PWCNet and VGG19 but does not provide specific ancillary software details like library or solver names with version numbers needed to replicate the experiment.
Experiment Setup Yes In compound regularization, W( ) is implemented by warping with a random optical flow, while Δ is a random noise with Δ N(0, σ2I), σ2 U(0.01, 0.02). The network is first pre-trained without Lt for two epochs, then fine-tuned with Lt for 5k iterations. To alleviate this problem, we set λt = 150 so that with Ls = 1.067207, the stylization effect is still pleasing.