Augmented Balanced Image Dataset Generator Using AugStatic Library
The mixed data consists of various structured and unstructured data. The exponential boom of the amount of data has made the datasets of varying samples. This paper focuses on the image dataset generator that balances an imbalanced dataset using the AugStatic augmentation library. The datasets, including various classes, are said to be balanced if the number of samples in the classes is equal. This gives a fair chance for the model to learn about all the classes. An augmented image dataset balances an imbalanced image dataset. It is useful when the data is less in a specific category, generating new data with it. There are multiple augmentation techniques supported by the AugStatic library that helps in developing the augmented balanced library by the iterative implementation of the augmentations on the generated dataset. It takes an input of the existing dataset, majority and minority classes sample count that returns a balanced image dataset by iteratively applying the augmentations on the generated augmented images in the minority class. This generator is efficient and can be used for any image dataset.
Code (1)
Tasks
ClassificationData AugmentationData VisualizationDetecting Image ManipulationImage AugmentationImage ClassificationImage CroppingImage DenoisingImage GenerationImage ManipulationImage Manipulation DetectionImage MattingImage Morphingimage smoothingImage StitchingImage-VariationROI-based image generationMethods 이 논문이 사용한 방법론
Similar Papers 제목 키워드 기반
AugStatic - A Light-Weight Image Augmentation Library
The rapid exponential increase in the data led to an abrupt mix of various data types, leading to a deficiency of helpful information. Creating new data with the existing different types of data are presented in this pap…
ClassificationData AugmentationData VisualizationDetecting Image Manipulation+13Unbalanced GANs: Pre-training the Generator of Generative Adversarial Network using Variational Autoencoder
We propose Unbalanced GANs, which pre-trains the generator of the generative adversarial network (GAN) using variational autoencoder (VAE). We guarantee the stable training of the generator by preventing the faster conve…
Generative Adversarial NetworkSynthetic Augmentation in Imbalanced Learning: When It Helps, When It Hurts, and How Much to Add
Imbalanced classification often causes standard training procedures to prioritize the majority class and perform poorly on rare but important cases. A classic and widely used remedy is to augment the minority class with …
Linguistically Augmented Audio Speech Data (LinguAS)
Maliciously-created fake speech, including deepfaked and spoofed audio, is proliferating at an alarming rate, and detection models are racing to stay ahead of the curve. Yet, most detection models are trained to make inf…
Feature-Augmented Transformers for Robust AI-Text Detection Across Domains and Generators
AI-generated text is nowadays produced at scale across domains and heterogeneous generation pipelines, making robustness to distribution shift a central requirement for supervised binary detectors. We train transformer-b…
Text Detection