Guiding the One-to-One Mapping in CycleGAN via Optimal Transport

Authors: Guansong Lu, Zhiming Zhou, Yuxuan Song, Kan Ren, Yong Yu4432-4439

AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments indicate that the proposed algorithm is capable of learning a one-to-one mapping with the desired properties. We conduct several image-to-image translation experiments between different datasets, and we compare the translation results of our algorithm with Cycle GAN.
Researcher Affiliation Academia Guansong Lu, Zhiming Zhou, Yuxuan Song, Kan Ren, Yong Yu Shanghai Jiao Tong University {gslu, heyohai, songyuxuan, kren, yyu}@apex.sjtu.edu.cn
Pseudocode No The paper does not contain explicitly labeled pseudocode or algorithm blocks. The algorithm is described in prose within the 'Guiding Cycle GAN with Optimal Transport' section.
Open Source Code No The paper does not provide any specific links to source code or an explicit statement about its release.
Open Datasets Yes We conduct our first experiment between a car dataset (Fidler, Dickinson, and Urtasun 2012) and a chair dataset (Aubry et al. 2014). In this experiment, we performed image-to-image translation between a shoes dataset (Yu and Grauman 2014) and a handbags dataset (Zhu et al. 2016).
Dataset Splits No The paper mentions training models but does not specify the explicit splits for training, validation, or testing sets. It refers to datasets but not how they were partitioned for reproducibility.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, or memory specifications) used for running the experiments.
Software Dependencies No The paper mentions using 'network simplex algorithm' and 'Adam optimizer' but does not specify version numbers for these or any other software dependencies.
Experiment Setup Yes We train our model for 3000 epochs with an initial learning rate of 0.0002 and linearly decayed it to zero. λgp is set as 10, λrec is set in the range of [100, 800] and λref is set in the range of [50, 300]. We train critic for 5 steps and generator for 1 step in turn.