Controlling generative models with continuous factors of variations

Authors: Antoine Plumerault, Hervé Le Borgne, Céline Hudelot

ICLR 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate the effectiveness of our method qualitatively and quantitatively, both for GANs and variational auto-encoders. and 3 EXPERIMENTS Datasets: We performed experiments on two datasets.
Researcher Affiliation Academia Antoine Plumerault , Hervé Le Borgne , Céline Hudelot CEA, LIST, Laboratoire Analyse Sémantique Texte et Image, Gif-sur-Yvette, F-91191 France Université Paris-Saclay, Centrale Supélec, 91190, Gif-sur-Yvette, France.
Pseudocode Yes Algorithm 1: Create a dataset of trajectories in the latent space which corresponds to a transformation T in the pixel space.
Open Source Code Yes All our experiments have been implemented with Tensor Flow 2.0 (Abadi et al., 2015) and the corresponding code is available on github here.
Open Datasets Yes Datasets: We performed experiments on two datasets. The first one is d Sprites (Matthey et al., 2017), composed of 737280 binary 64 64 images containing a white shape on a dark background. The second dataset is ILSVRC (Russakovsky et al., 2015), containing 1.2M natural images from one thousand different categories.
Dataset Splits No The paper refers to
Hardware Specification No The paper does not provide specific hardware details such as GPU models, CPU types, or memory used for running the experiments.
Software Dependencies Yes All our experiments have been implemented with Tensor Flow 2.0 (Abadi et al., 2015)
Experiment Setup Yes The models were trained on d Sprites (Matthey et al., 2017) with an Adam optimizer during 1e5 steps with a batch size of 128 images and a learning rate of 5e 4.