NaturalInversion: Data-Free Image Synthesis Improving Real-World Consistency

Authors: Yujin Kim, Dogyun Park, Dohee Kim, Suhyun Kim1201-1209

AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we evaluate the performance of Natural Inversion on CIFAR-10 and CIFAR-100 (Krizhevsky and Hinton 2009). Our experiments contain two parts, (1) analysis of our method: we verify that our method can synthesize more natural images (2) applications: we ensure the effectiveness of Natural Inversion on various applications in data-free conditions.
Researcher Affiliation Academia Yujin Kim1,2, Dogyun Park1,2, Dohee Kim2, Suhyun Kim2* 1Korea University, Republic of Korea 2Korea Institute of Science and Technology, Republic of Korea
Pseudocode Yes Algorithm 1: Natural Inversion Algorithm
Open Source Code No The paper does not provide an explicit statement or a link to open-source code for the described methodology.
Open Datasets Yes We evaluate the performance of Natural Inversion on CIFAR-10 and CIFAR-100 (Krizhevsky and Hinton 2009).
Dataset Splits No The paper mentions 'Our experiments settings can be found in appendix' but does not provide specific details on validation splits within the main text.
Hardware Specification No The paper does not specify any hardware details such as GPU models, CPU types, or cloud computing resources used for the experiments.
Software Dependencies No The paper does not provide specific version numbers for any software dependencies or libraries used in the implementation.
Experiment Setup Yes We set the mini-batch size as 128 and train the model for 200 epochs using an SGD optimizer with a 0.05 learning rate.