Flow-based Image-to-Image Translation with Feature Disentanglement
Authors: Ruho Kondo, Keisuke Kawano, Satoshi Koide, Takuro Kutsuna
NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We compared our model to the official implementations of VUNet [1] and PUNet [2]. Datasets We employed the celebrity face attributes dataset (Celeb A [19], which consists of images with 40 attribute annotations) as well as our orginal dataset, Cahn Hilliard Cook (CHC) dataset. Table 1: Comparison of FID, LPIPS and c LPIPS. The means and standard deviations of five trials are shown for FID. |
| Researcher Affiliation | Industry | Ruho Kondo Toyota Central R&D Labs. r-kondo@mosk.tytlabs.co.jp Keisuke Kawano Toyota Central R&D Labs. kawano@mosk.tytlabs.co.jp Satoshi Koide Toyota Central R&D Labs. koide@mosk.tytlabs.co.jp Takuro Kutsuna Toyota Central R&D Labs. kutsuna@mosk.tytlabs.co.jp |
| Pseudocode | Yes | Algorithm 1 Procedures used to obtain condition-specific diverse images xsp, condition-invariant diverse images xiv, and style transformed images xtrans. ( )mean is the mean of ( ). |
| Open Source Code | No | The paper does not contain an explicit statement or link providing access to the open-source code for the described methodology. |
| Open Datasets | Yes | Datasets We employed the celebrity face attributes dataset (Celeb A [19], which consists of images with 40 attribute annotations) as well as our orginal dataset, Cahn Hilliard Cook (CHC) dataset. |
| Dataset Splits | Yes | The number of data for training, validation and testing in Celeb A were 162,770, 1,9867 and 19,962, respectively, which follows an official train/val/test partitions. |
| Hardware Specification | Yes | All experiments were carried out on a single NVIDIA Tesla P100 GPU. |
| Software Dependencies | Yes | We implemented our model with Tensor Flow version 1.10.0 [30]. |
| Experiment Setup | Yes | In all of our experiments, α = 0.01 and β = 0.1 were used. For training, we used the default settings of the Adam [32] optimizer with a learning rate of 10 4 and batch size of 32. |