This repository contains links to two tools, namely [multi-label noise injection](https://gitlab.tubit.tu-berlin.de/rsim/multi_label_noise) and [noisifier](https://gitlab.tubit.tu-berlin.de/rsim/noisifier), which both help to inject label noise into image datasets. The resulting noisified dataset can be used in experiments to train machine learning models robust against label noise. Both tools work on [Numpy](https://numpy.org/) arrays and are independent of specific deep learning libraries.
This page contains links to the tools developed at the [Remote Sensing Image Analysis group](https://www.rsim.tu-berlin.de/menue/remote_sensing_image_analysis_group/) to inject synthetic multi-label noise into image datasets. The resulting noisy labeled datasets can be used in experiments to evaluate the robustness of the machine learning models against label noise. Both tools work on [Numpy](https://numpy.org/) arrays and are independent of specific deep learning libraries.
## 1) multi-label noise injection
[multi-label noise injection](https://gitlab.tubit.tu-berlin.de/rsim/multi_label_noise) contains a set of helper functions in order to create noisy multi-label matrices. The tool allows to inject additive (extra classes) and subtractive (missing classes) noise separately.
## 1) Multi Label Noise Injection Tool
[Multi-Label Noise Injection tool](https://gitlab.tubit.tu-berlin.de/rsim/multi_label_noise) contains a set of helper functions in order to create noisy multi-label matrices. The tool allows to inject additive (extra classes) and subtractive (missing classes) noise separately.
[noisifier](https://gitlab.tubit.tu-berlin.de/rsim/noisifier) allows to add noise to the labels of a dataset. The dataset can be singlelabel or multi-label.
## 2) Noisifier Tool
[Noisifier tool](https://gitlab.tubit.tu-berlin.de/rsim/noisifier) allows to add synthetic multi-label noise to the multi-labeled image datasets. It is also applicable to the datasets containing single-labeled images.
