Abstract : Deep learning methodologies have been used to create applications with utility in fields such as computer vision, natural language processing, speech recognition, etc. To train such neural networks that work on these specific tasks, the process of the collection of data becomes paramount. As massive datasets become more prevalent, the models need to be unbiased and detect the minority of classes in the dataset without population bias. To mitigate this problem, the Cycle Generative Adversarial Neural networks can be utilized. Depending on the datasets, one can determine whether minority classes would lie in space where they could be transformed into the different class. For image datasets CycleGANs can be used for image synthesis.

Keywords: Generative Adversarial Networks, CycleGANs, Unbalanced distribution of classes, Deep learning, Synthesis

| DOI: 10.17148/IARJSET.2022.9633