Exemplar Guided Unsupervised Image-to-Image Translation with Semantic Consistency

Authors: Liqian Ma, Xu Jia, Stamatios Georgoulis, Tinne Tuytelaars, Luc Van Gool

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results on various datasets show that EGSC-IT does not only translate the source image to diverse instances in the target domain, but also preserves the semantic consistency during the process.
Researcher Affiliation Collaboration 1KU-Leuven/PSI, TRACE (Toyota Res in Europe) 2KU-Leuven/PSI 3ETH Zurich 4Huawei Noah s Ark Lab
Pseudocode No The paper describes the framework and learning procedures but does not include a formal pseudocode or algorithm block.
Open Source Code No The paper does not provide an explicit statement or link indicating the availability of its source code.
Open Datasets Yes We set up a controlled experiment on the MNIST-Single dataset, which is created based on the handwritten digits dataset MNIST Le Cun et al. (1998)... We carry out a synthetic real experiment for street view translation between GTA5 (Richter et al., 2016) and Berkeley Deep Drive (BDD) (Xu et al., 2017) datasets... The Large-scale Celeb Faces Attributes (Celeb A) dataset (Liu et al., 2015)... we use EMNIST (Cohen et al., 2017)
Dataset Splits No The MNIST-Single dataset consists of two different domains as shown in Fig. 4. For domain A of both training/test sets, the foreground and background are randomly set to black or white but different from each other. For domain B of training set, the foreground and background for digits from 0 to 4 are randomly assigned a color from red, green, blue , and the foreground and background for digits from 5 to 9 are fixed to red and green, respectively. For domain B of testing set, the foreground and background of all digits are randomly assigned a color from red, green, blue .
Hardware Specification No The paper does not provide specific hardware details such as GPU or CPU models used for the experiments.
Software Dependencies No The paper mentions optimizers (Adam) and neural network architectures (VGG, Deeplab) but does not provide specific version numbers for software dependencies or libraries.
Experiment Setup Yes Table 6: Network architecture and training parameters. Translation n1 n2 n3 Minibatch Learning rate λs λc Iteration Single-digit 1 4 5 8 1e-5 1e3 1e1 60k Multi-digit 3 4 5 8 1e-5 1e4 1e2 60k GTA5 BDD 3 4 5 3 1e-4 1e4 1e2 22k Face gender 3 4 5 8 1e-4 5e3 1e1 30k. We use the Adam Kingma & Ba (2015) optimizer with β1 = 0.5 and β2 = 0.999. The learning rate is polynomially decayed with a power of 0.9...