Incorporating long-range consistency in CNN-based texture generation

Authors: Guillaume Berger, Roland Memisevic

ICLR 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 4 EXPERIMENTS
Researcher Affiliation Academia Guillaume Berger & Roland Memisevic Department of Computer Science and Operations Research University of Montreal guillaume.berger@umontreal.ca, memisevr@iro.umontreal.ca
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code Yes Our implementation4 uses Lasagne (Dieleman et al., 2015).4available at https://github.com/guillaumebrg/texture generation
Open Datasets No Most textures used as references in this paper were taken from textures.com and pixabay.com. This does not provide concrete access information for a publicly available or open dataset.
Dataset Splits No No specific dataset split information (percentages, counts, or predefined splits) for training, validation, or testing was provided.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory, or cloud instance types) used for running experiments.
Software Dependencies No Our implementation4 uses Lasagne (Dieleman et al., 2015). The paper mentions "Lasagne" but does not provide a specific version number.
Experiment Setup Yes for image sizes of roughly 384 384 pixels, we recommend the following δ values per layer (which we used in all our following experiments): {2, 4, 8, 16, 32, 64} for pool1, {2, 4, 8, 16, 32} for pool2, {2, 4, 8, 16} for pool3, and {2, 4, 8} for pool4.