Code for the paper, "Distribution Augmentation for Generative Modeling", ICML 2020.

Official implementation of the paper "Diffusion-based Image Generation for In-distribution Data Augmentation in Surface Defect Detection" accepted @ VISAPP 2024.

Official implementation of the paper "Leveraging Latent Diffusion Models for Training-Free In-Distribution Data Augmentation for Surface Defect Detection" accepted @ CBMI 2024.

The official implementation of paper "Unsupervised Few-Shot Learning via Distribution Shift-based Augmentation"

This is a repo of Out-of-Distribution Generalization via Decomposed Feature Representation and Semantic Augmentation.

[NeurIPS 2023] "Unleashing the Power of Graph Data Augmentation on Covariate Distribution Shift" by Yongduo Sui, Qitian Wu, Jiancan Wu, Qing Cui, Longfei Li, Jun Zhou, Xiang Wang, Xiangnan He.

Official PyTorch implementation of the paper "IPA: Image Patch Augmentation via Repeated Rearrangement for Zero-Shot Out-of-Distribution Detection"

DistAwareAug introduces a new paradigm: statistically-governed augmentation. It learns the statistical distribution of minority classes (mean, variance, covariance, correlations) and generates synthetic samples that preserve these properties. A distance-aware mechanism ensures diversity while avoiding unrealistic outliers or duplicates.

The implementation of Paper: Dancing with Noise: Advancing Generative Speech Enhancement with Distribution Augmentation.

Official repository for ECCV2024 publication, TAG: Text Prompt Augmentation for Zero-Shot Out-of-Distribution Detection

Code of the NeurIPS 2022 paper "Hyperbolic Feature Augmentation via Distribution Estimation and Infinite Sampling on Manifolds".

CALM-Aug is a lightweight, class-aware data augmentation framework designed to alleviate long-tail distribution and domain shift problems in agricultural object detection datasets such as PlantDoc.

Distribution-preserving data-augmentation

AugmentationLab explores the role of Data Augmentation in Robustness, Class Imbalance, Distribution Shift and Generalization.

Distributed Distributional DrQ is a model-free and off-policy RL algorithm for continuous control tasks based on the image-vision input. This is an actor-critic method with data-augmentation and D4PG as the backbone.

A specialized toolkit for industrial time-series data augmentation and quantitative evaluation—focusing on core issues such as data scarcity, poor quality, and distribution shifts in industrial time-series data. It enhances data support for developing highly reliable industrial AI models .

The project shows implementation based on a variety of applications of convolutional networks. DCGAN structure is used for model construction on GAN to learn deeply with multiple convolutional layers. The accuracy for classification task is about doubled after being supported by GAN-based data augmentation. This quantity is even significantly increased with the suitable selection of batch size, which is distinctly differentiated with 2 given label categories. Particularly, by being provided by more high-quality synthesized images, CNN for cow can obtain up to 95% accuracy with the adequate batch size for the training process on GAN. Lastly, the application of surface-feature extraction of PatchGAN trained along with CycleGAN considering the cycle consistency loss of image reconstruction from a distribution to the target one helps grasping the mapping between them and creates generators able to synthesize the transferred version containing the style feature of the objective distribution from the image of the original one without requiring any paired supportive similarity.

No description available

Implicit Augmentation from Distributional Symmetry in Turbulence Super-Resolution

code for paper "An Efficient Data Augmentation Network for Out-of-Distribution Image Detection"

Codes for reproducing "Improving Transferability of Adversarial Examples by Saliency Distribution and Data Augmentation".

[ECML 2022] Official code for the paper "A Novel Data Augmentation Technique for Out-of-Distribution Sample Detection using Compounded Corruptions"

This repository contains the official implementation of "Data-Agnostic Augmentations for Unknown Variations - Out-of-Distribution Generalisation in MRI Segmentation", under review at MIDL 2025.

This research extends a pre-trained trajectory prediction model to enhance its ability to handle Out-of-Distribution (OOD) and Non-IID data in General Aviation (GA) flight trajectories. Key improvements include synthetic data augmentation, uncertainty estimation, and model fine-tuning.

No description available

Image augmentation tool for out-of-distribution (OOD) robustness evaluation implemented in C++ for fast augmentation of large image datasets.

Benchmarking the Stack Foundation Model for out-of-distribution cancer drug response prediction with Evo 2 genotype augmentation

Novel technique to extract a noise distribution from passive acoustic monitoring (PAM) arrays. Applications inlcude training denoising autoencoders and data augmentation.

Implementation of data augmentation by using PCA to 2D points, therefore dividing two-dimensional data into sections and subsections considering radious and angle of section, followed by sample generation based on density distribution across the entire plane

Use a neural network to classify an image of an eye to multiple levels of illness of the disease diabetic retinopathy. Used is the dataset ’Indian Diabetic Retinopathy Image Dataset’ (IDRID). To combat the imbalanced class distribution, over-, undersampling, and weighted loss were used. Implemented transfer learning to use more complex models without training them completly from scratch. Because of the small number of samples in the IDRID dataset we implemented data augmentation to create additional samples by modifying original sample images.