Hackathon : Completed #supervised_learning

πŸŽ“ This challenge is a part of the Application of Machine Learning in Plant Sciences Summer School

πŸ•΅οΈ Introduction

Plant diseases and pests affect the normal growth of plants and even cause plant death during the growth process. Crop diseases can become a major threat to food security, but their rapid identification remains difficult in many parts of the world due to the lack of the necessary infrastructure. But with easy access to phones and cameras, these diseases can be captured and identified to avoid plant death and improve crop yield. 


Can you detect plant diseases just by examining the images of the plant’s leaves, even when the resolution is low? Participants must correctly identify a disease class from low-resolution images of plant leaves in this task

Get started with this easy-2-follow notebook

πŸ’Ύ Dataset

The dataset consists of 43525 images of plants leave at 32x32 pixels. Each of the images belongs to one of the 38 disease classes (some of these classes also represent healthy leaves). A separate test dataset of 10779 images is provided, and the task is to predict the associated disease (from the 38 valid disease classes) for each of the images. The training and the test set reflect numerous real-world properties of datasets, such as class imbalance (some classes have about ~4500 images, while some classes can have as low as ~120 images in the training set) etc.

The images for both the training and the test set are aggregated into individual Numpy array files for convenient programmatic access.

πŸ“ Files

The following files are available in the resources section:

  • train-images.npy - (43466 samples) A numpy file with an ndarray of shape (43525, 32, 32, 3) representing all the images in the training set.
  • train-labels.npy - (43466 samples) A numpy file with a ndarray shape (43525,) representing the corresponding labels for the images in the training set. This array consists of a unique index per class. And mapping of the index to a human-readable class name can be found in all_classes.txt.
  • test-images.npy - (10838 samples) A numpy file with an ndarray of shape (10779, 32, 32, 3) representing all the images in the test set.
  • all_classes.txt - A file containing the mapping of the class index used in the train-labels.npy file, and the sample_submission.csv file to that of a human-readable class name. The file contains 38 lines, where the line index of the class name in the file is the index of the class names in all the label files.
  • sample_submission.csv - A sample submission file to provide a reference about the expected file format for the submission system to work.

πŸš€ Submission

  • Prepare a CSV containing a header as class_index and the predicted class index [0-38) with the name as submission.csv.
  • Sample submission format available as sample_submission.csv in the resources section.

Make your first submission here πŸš€ !!
Explore the Getting Started Notebook here πŸš€ !!

πŸ“• Resources 

  1. Plant diseases and pests detection based on deep learning
  2. Using Deep Learning for Image-Based Plant Disease Detection
  3. Plant Disease Detection by Imaging Sensors – Parallels and Specific Demands for Precision Agriculture and Plant Phenotyping
  4. Plant leaf disease detection using computer vision and machine learning algorithms

πŸ–Š Evaluation Criteria 

During the evaluation, the F1 score and Accuracy Score will be used to test the efficiency of the model where,


🎁 Prizes

NOTE: Only participants of the Application of Machine Learning in Plant Sciences Summer School 2022  are eligible for the prizes. 




See all