PlantWild Dataset
 
    Imagic: Text-Based Real Image Editing with Diffusion Models

    Benchmarking In-the-Wild Multimodal Plant Disease Recognition and A Versatile Baseline

    Tianqi Wei Zhi Chen Zi Huang Xin Yu

    arXiv  |  GitHub  |  Downloads




    Introduction

    Existing deep-learning methods have achieved remarkable performance in recognizing in-laboratory plant disease images. However, their performance often significantly degrades in classifying in-the-wild images. Furthermore, we observed that in-the-wild plant images may exhibit similar appearances across various diseases (i.e., small inter-class discrepancy) while the same diseases may look quite different (i.e., large intra-class variance). Motivated by this observation, we propose an in-the-wild multimodal plant disease recognition dataset, PlantWild, which contains the largest number of disease classes but also text-based descriptions for each disease. PlantWild is currently the largest dataset containing wild plant disease images.

    Our PlantWild dataset can be regarded as an ideal testbed for evaluating disease recognition methods in the real world. The dataset is accessible through Download Page.

    Method

    Code of our baseline MVPDR is available through https://github.com/tqwei05/MVPDR. The workflow of MVPDR is presented as the following figure. It models text descriptions and visual data through multiple prototypes and can achieve outstanding performance on in-the-wild plant disease images.

    CLIP encoders extract features from images and text for each category and then multiple prototypes are constructed by grouping visual features. Given a test image, both the visual and textual prototypes can be used for classification. Therefore, MVPDR can not only classify diseases but also recognize diseases in few-shot or training-free scenarios.

    Paper

    "Benchmarking In-the-Wild Multimodal Plant Disease Recognition and A Versatile Baseline",

    Tianqi Wei, Zhi Chen, Zi Huang, Xin Yu.

    ACM International Conference of Multimedia 2024

    [PDF]

    Downloads

    PlantWild

    The PlantWild dataset is accessible through:

    PlantWild_v2

    We invited experts in the field of agriculture to refine our dataset. We also added images of new types of diseases according to experts' suggestions. The obtained dataset, PlantWild_v2, has enhanced data reliability and the number of classes has been expanded to 115.
    This version can be accessed through:

    PlantSeg with Segmentation annotations

    On top of the plant disease recognition dataset PlantWild, we present the PlantSeg dataset that provides segmentation annotations focused on diseased parts for images.

    Copyright

    Our dataset follows the copyright Creative Commons BY-NC-ND 4.0 license.

    BibTeX

    
  @inproceedings{MVPDR,
  title={Benchmarking In-the-Wild Multimodal Plant Disease Recognition and A Versatile Baseline},
  author={Wei, Tianqi and Chen, Zhi and Huang, Zi and Yu, Xin},
  booktitle={ACM International Conference of Multimedia},
  year={2024}}

    Search