UAPS

KO | EN

gitlite — search

#javascript #python #hacktoberfest #react #ai #typescript #llm #go #golang #android #machine-learning #rust #deep-learning #linux

(★ 14)

This repo contains implementation of uncertainty estimation, rectification, and minimization for guiding the pseudo-label learning in semi-supervised defect segmentation setting.

UAPS Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Uncertainty-aware and Dynamically-mixed Pseudo-labels for Semi-supervised Defect Segmentation (UAPS)
## 📌 Overview
In this study, we propose a novel uncertainty-aware pseudo labels, which are generated from dynamically mixed predictions of multiple decoders that leverage a shared encoder network. The estimated uncertainty guides the pseudo-label-based supervision and regularizes the training when using the unlabeled samples. In our experiments on four public datasets for defect segmentation, the proposed method outperformed the fully supervised baseline and six state-of-the-art semi-supervised segmentation methods. We also conducted an extensive ablation study to demonstrate the effectiveness of our approach in various settings.

## 🚀 Full paper source:
You can read the details about the methods, implementation, and results from: (https://www.sciencedirect.com/science/article/pii/S0166361523001458)

⭐ [Please Star this repo](https://github.com/djene-mengistu/UAPS)

🔥 If you find it useful, **also cite our work as follows:**
```
@article{SIME2023103995,
title = {Uncertainty-aware and dynamically-mixed pseudo-labels for semi-supervised defect segmentation},
journal = {Computers in Industry},
volume = {152},
pages = {103995},
year = {2023},
issn = {0166-3615},
doi = {https://doi.org/10.1016/j.compind.2023.103995},
url = {https://www.sciencedirect.com/science/article/pii/S0166361523001458},
author = {Dejene M. Sime and Guotai Wang and Zhi Zeng and Bei Peng},
}
```
## 🔥 Challenges
The illustration of the uncertainty map and error is presented as follows:

<img src="/fig_data/uncertainty.jpg" width="40%" height="10%">

**Fig. 1:** Illustration of the network prediction, prediction error which is false negatives and false positives, and the uncertainty map computed as the KL-distance between primary
decoder and average predictions.

## 🏗️ The UAPS Frameworks 
The overall architecture of the proposed method is presented as follows:

<img src="/fig_data/main-arch.jpg" width="60%" height="40%">

**Fig. 2:** Illustration of the overall framework of the UAPS using the unlabeled samples. We apply various perturbations to the encoder output to obtain different predictions with multiple decoders. The predictions from the decoders are then combined using randomly generated weights to compute a pseudo-label. We calculate uncertainty as the KL-distance between the average prediction and each decoder’s prediction.

The baseline architecture and the perterbation method is presented as follows:

<img src="/fig_data/UNET-baseline.jpg" width="60%" height="40%">

**Fig. 3:** Illustration of baseline architecture adapted from U-Net with perturbation module. The encoder output remains unchanged at each block, while the type of perturbation changes to produce different versions of the outputs to be fed to the decoder networks.

## 🧠 Data Preparation
Download datasets. There are 4 datasets to download:
* NEU-SEG dataset from [NEU-seg](https://ieeexplore.ieee.org/document/8930292)
* DAGM dataset from [DAGM](https://www.kaggle.com/datasets/mhskjelvareid/dagm-2007-competition-dataset-optical-inspection)
* MT (Magnetic Tiles) dataset from [MTiles](https://www.kaggle.com/datasets/alex000kim/magnetic-tile-surface-defects)
* KoSDD2 (KolektorSDD2) dataset from [KoSDD2](https://www.vicos.si/resources/kolektorsdd2/)

Put downloaded data into the following directory structure:
* .../data/
    * NEU_data/ ... # raw data of NEU-Seg
    * DAGM_data/ ...# raw data of DAGM
    * MTiles_data/ ...# raw data of MTiles
    * KoSDD2_data/ ...# raw data of KoSDD2
## 📌 Code usage
The training files and settings for each compared network is presented in separate directory. Train each network and test from the presented directory.
To train the proposed **UAPS** method run the following after setting hyperparameters such as labeled-ratio, iteration-per-epoch, consistency ramp length, and consistency loss coefficients.
```bash
python UAPS_train.py
```

To test the performance of the proposed method, load the trained weights accordingly and run the following:
```bash
run UAPS_Testing.ipynb
```

Similarly, train the proposed method, **UAPS**, for the other datasets from the indicated directories after setting appropriate hyper-parametres.

## 🌱 Some results and visualizations
Visualization of the segmetnation results on selected samples are presented as follows:

<img src="/fig_data/neu-viz.jpg" width="60%" height="50%">

**Fig. 4:** Visualization of the segmentation results on the NEU-Seg dataset. The regions indicated by the dashed-red-box shows wrong prediction.

**Fig. 5:** Visualization of the segmentation results on the MTiles dataset.

## 🤝 Acknowledgments

This repo borrowed many implementations from [simEps](https://github.com/djene-mengistu/simEps)

## 📬 Contact
For any issue please contact me at djene.mengistu@gmail.com

UAPS

UAPS Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

UAPS

UAPS Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README