Multi-View Data Generation Without View Supervision
Authors: Mickael Chen, Ludovic Denoyer, Thierry Artières
ICLR 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We experiment it on four image datasets on which we demonstrate the effectiveness of the model and its ability to generalize. |
| Researcher Affiliation | Collaboration | Micka el Chen Sorbonne Universit e, CNRS, Laboratoire d Informatique de Paris 6, LIP6, F-75005, Paris, France mickael.chen@lip6.fr Ludovic Denoyer Sorbonne Universit e, CNRS, Laboratoire d Informatique de Paris 6, LIP6, F-75005, Paris, France Criteo Research ludovic.denoyer@lip6.fr Thierry Arti eres Aix Marseille Univ, Universit e de Toulon, CNRS, LIS, Marseille, France Ecole Centrale Marseille thierry.artiere@centrale-marseille.fr |
| Pseudocode | No | The paper describes the model architecture and objective functions but does not include any explicit pseudocode or algorithm blocks. |
| Open Source Code | Yes | Following the article, an implementation of our algorithms is freely available1. 1https://github.com/mickaelchen/GMV |
| Open Datasets | Yes | Celeb A (Liu et al. (2015)) 198791 3808 9999 178 1 19.9 35 3DChairs (Aubry et al. (2014)) 80600 5766 1300 93 62 62 62 MVC cloth (Liu et al. (2016)) 159128 2132 37004 495 4 4.3 7 102flowers (Nilsback & Zisserman (2008)) 8189 102 40 80.3 258 |
| Dataset Splits | No | Table 1 lists 'train' and 'test' data splits for the datasets, but no explicit 'validation' split is mentioned with specific percentages or counts. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU models, CPU types, memory) used for running the experiments. |
| Software Dependencies | No | The paper mentions using Adam optimizer and DCGAN implementation, but it does not provide specific version numbers for any software dependencies or libraries (e.g., Python, PyTorch, TensorFlow, specific GAN frameworks). |
| Experiment Setup | Yes | The images were rescaled to 3 64 64 tensors. The generator G and the discriminator D follow that of the DCGAN implementation proposed in Radford et al. (2015). Learning has been made using classical GAN learning techniques: we used Adam optimizer (Kingma & Ba (2014)) with batches of size 128. Following standard practice, learning rate in the GMV experiments are set to 1 10 3 of G and 2 10 4 for D. For the C-GMV experiments, learning rates are set to 5 10 5. The adversarial objectives are optimized by alternating gradient descent over the generator/encoder, and over the discriminator. |