Beyond-supervision-Harnessing-self-supervised-learning-in-unseen-plant-disease-recognition

Introduction

This is a pytorch implementation for CL-ViT and FF-ViT in Beyond supervision: Harnessing self-supervised learning in unseen plant disease recognition

Proposed CL-ViT architecture.

Proposed FF-ViT architecture.

The contributions of this paper:

1. We demonstrate that the incorporation of a guided learning mechanism surpasses conventional approaches in the multi-plant disease identification benchmark. Furthermore, we show that the CL-ViT model, integrating a SSL approach, outperforms the FF-ViT model employing a purely supervisory learning scheme for unseen plant disease identification tasks.
2. In our qualitative analyses, we illustrate that CL-ViT learns a feature space capable of discriminating between different classes while minimizing the domain gap between seen and unseen data. This underscores the superiority of CL-ViT in implementing a more effective guided learning mechanism.

Results

Preparation

• Dataset: spMohanty Github
  (You can group all images into single folder to directly use the csv file provided in this repo)

• download ViT pretrained weight link (From rwightman Github timm repo)

Implementations

CL-ViT >> code

Notes

• The csv file (metadata of images) are here

FF-ViT >> code

Notes

• The csv file (metadata of images) are here
  (path_list.csv to locate the csv.file for all crop and disease training classes)

See also

1. Pairwise Feature Learning for Unseen Plant Disease Recognition: The first implementation of FF-ViT model with moving weighted sum. The current work improved and evaluated the performance of FF-ViT model on larger-scale dataset.
2. Unveiling Robust Feature Spaces: Image vs. Embedding-Oriented Approaches for Plant Disease Identification: The analysis between image or embedding feature space for plant disease identifications.

Dependencies

Pandas == 1.4.1
Numpy == 1.22.2
torch == 1.10.2
timm == 0.5.4
tqdm == 4.62.3
torchvision == 0.11.3
albumentations == 1.1.0

License

Creative Commons Attribution-Noncommercial-NoDerivative Works 4.0 International License (“the CC BY-NC-ND License”)

Citation

@article{chai2024beyond,
  title={Beyond supervision: Harnessing self-supervised learning in unseen plant disease recognition},
  author={Chai, Abel Yu Hao and Lee, Sue Han and Tay, Fei Siang and Bonnet, Pierre and Joly, Alexis},
  journal={Neurocomputing},
  pages={128608},
  year={2024},
  publisher={Elsevier}
}