High-Fidelity Image Generation With Fewer Labels
Authors: Mario Lučić, Michael Tschannen, Marvin Ritter, Xiaohua Zhai, Olivier Bachem, Sylvain Gelly
ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct an extensive comparison of methods detailed in Table 1, namely: Unmodified BIGGAN, the unsupervised methods SINGLE LABEL, RANDOM LABEL, CLUSTERING, and the semi-supervised methods S2GAN and S2GAN-CO. In all S2GAN-CO experiments we use soft labels... |
| Researcher Affiliation | Collaboration | 1Google Research, Brain Team 2ETH Zurich. |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks that are clearly labeled 'Pseudocode' or 'Algorithm'. |
| Open Source Code | Yes | open-source all the code used for the experiments at github.com/google/compare_gan. |
| Open Datasets | Yes | Datasets We focus primarily on IMAGENET, the largest and most diverse image data set commonly used to evaluate GANs. IMAGENET contains 1.3M training images and 50k test images, each corresponding to one of 1k object classes. |
| Dataset Splits | Yes | Partially labeled data sets for the semi-supervised approaches are obtained by randomly selecting k% of the samples from each class. The parameters γ and B are tuned on 0.1% labeled examples held out from the training set, the search space is {0.1, 0.5, 1.0} × {1024, 1536, 1792}. |
| Hardware Specification | Yes | All models are trained on 128 cores of a Google TPU v3 Pod with Batch Norm statistics synchronized across cores. |
| Software Dependencies | No | The paper mentions software components like 'Adam optimizer' and 'Batch Norm' but does not provide specific version numbers for any software dependencies or libraries. |
| Experiment Setup | Yes | We use exactly the same optimization hyper-parameters as Brock et al. (2019). Specifically, we employ the Adam optimizer with the learning rates 5 × 10−5 for the generator and 2 × 10−4 for the discriminator (β1 = 0, β2 = 0.999). We train for 250k generator steps with 2 discriminator steps before each generator step. The batch size was fixed to 2048, and we use a latent code z with 120 dimensions. |