ERA is in the process of being migrated to Scholaris, a Canadian shared institutional repository service (https://scholaris.ca). Deposits to existing ERA collections are frozen until migration is complete. Please contact erahelp@ualberta.ca for further assistance
- 268 views
- 139 downloads
Synthetic image data generation using GAN with statistical similiarity
-
- Author(s) / Creator(s)
-
Synthetic image generation using Generative Adversarial Networks (GANs) has emerged as a promising technique to address the challenge of limited datasets in the field of garbage classification. With supervised machine learning algorithms relying on labeled data and larger training examples, the small size and scarcity of annotated samples in the medical imaging domain pose significant obstacles. Traditional data augmentation techniques offer limited diversity, motivating the exploration of synthetic data examples to introduce more variability and enhance the training process. GANs, known for their ability to generate high-quality and realistic images, have gained popularity in computer vision tasks and have been successfully applied to medical imaging applications. In this study, we investigate the application of GANs for synthetic image generation in the garbage classification dataset. By training the GAN model with a large number of images, we aim to generate synthetic images that closely resemble real images, thereby expanding the dataset and improving the robustness of garbage classification algorithms. We also evaluate the similarity between
the real and generated images using metrics such as the Inception Score and Frechet Inception Distance. Through our research, we seek to demonstrate the efficacy of GAN-based synthetic image generation as a means to enhance the garbage classification dataset and improve the accuracy of classification algorithms for efficient waste management. -
- Date created
- 2023
-
- Subjects / Keywords
-
- Type of Item
- Research Material