Controlling Neural Style Transfer with Deep Reinforcement Learning
Authors: Chengming Feng, Jing Hu, Xin Wang, Shu Hu, Bin Zhu, Xi Wu, Hongtu Zhu, Siwei Lyu
IJCAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results demonstrate the effectiveness and robustness of our method. and Section 4: Experiments We have conducted a series of experiments to evaluate the effectiveness of RL-NST in realizing step-wise style transfer on both image and video NST tasks. |
| Researcher Affiliation | Collaboration | 1Chengdu University of Information Technology, China 2University at Buffalo, SUNY, USA 3Carnegie Mellon University, USA 4Microsoft Research Asia, China 5University of North Carolina at Chapel Hill, USA |
| Pseudocode | Yes | The pseudo-code of optimizing RL-NST is described in Algorithm 1. |
| Open Source Code | Yes | Our code, a user study, and additional results with more detailed information can be found in the supplementary materials. |
| Open Datasets | Yes | For image style transfer, we select style images from Wiki Art [Phillips and Mackintosh, 2011] and use MS-COCO [Lin et al., 2014] as content images in which the training set includes 80K images and the test set includes 40K images. and For video style transfer, we randomly collect 16 videos of different scenes from pexels[pex, 2022]... we use the training set of MPI Sintel dataset [Butler et al., 2012] as the test set... |
| Dataset Splits | No | The paper mentions training and test sets for datasets like MS-COCO (80K training, 40K test) and MPI Sintel (training set used as test set, 1K frames) but does not specify a separate validation set split or percentages for reproduction. |
| Hardware Specification | Yes | The speed is obtained with a Pascal Tesla P100 GPU. |
| Software Dependencies | No | The paper describes network architectures (CNN+RNN, Conv GRU, VGG) and mentions pre-trained models, but does not provide specific version numbers for software libraries, frameworks, or programming languages used for implementation. |
| Experiment Setup | Yes | In the experiment, we set λ = 1e5, β = 1e 7, ζ = 1e2 in Eq. (1), and η = 1e 4 in Eq. (2). |