Translate the Facial Regions You Like Using Self-Adaptive Region Translation

Authors: Wenshuang Liu, Wenting Chen, Zhanjia Yang, Linlin Shen2180-2188

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on three publicly available datasets, i.e. Morph, Ra FD and Celeb AMask-HQ, suggest that our approach demonstrate obvious improvement over state-of-the-art methods like Star GAN, SEAN and FUNIT.
Researcher Affiliation Academia Wenshuang Liu1,2,3, Wenting Chen1,2,3, Zhanjia Yang1,2,3, Linlin Shen1,2,3 1 Computer Vision Institute, School of Computer Science & Software Engineering, Shenzhen University 2 Shenzhen Institute of Artificial Intelligence & Robotics for Society 3 Guangdong Key Laboratory of Intelligent Information Processing {liuwenshuang2018, chenwenting2017, yangzhanjia2019}@email.szu.edu.cn, llshen@szu.edu.cn
Pseudocode No The paper describes the system framework and components but does not provide pseudocode or a clearly labeled algorithm block.
Open Source Code No The paper only provides a link to a video demo: '(https://youtu.be/DvIdmcR2LEc)', and does not explicitly state that the source code for the methodology is openly available or provide a link to a code repository.
Open Datasets Yes Our approach is extensively tested on three publicly available datasets, i.e. Morph (Ricanek and Tesafaye 2006), Ra FD (Langner et al. 2010) and Celeb AMask-HQ (Lee et al. 2020; Karras et al. 2017; Liu et al. 2015).
Dataset Splits No The paper specifies training and test set divisions for each dataset (e.g., 'We divide the images into a training set with 50020 images and a test set with 4,925 images' for Morph), but it does not explicitly state a separate validation dataset split.
Hardware Specification Yes We perform all traning runs on NVIDIA DGX with one Tesla V100 GPU using Pytorch 1.1.0 and cu DNN 7.4.2.
Software Dependencies Yes We perform all traning runs on NVIDIA DGX with one Tesla V100 GPU using Pytorch 1.1.0 and cu DNN 7.4.2.
Experiment Setup Yes Note that all the baselines are trained with the batch size of 4, the image size of 128 128 and the maximum iteration of 100,000.