icassava / fgvcx-icassava

Cassava disease prediction for the FGVCx competition of CVPR

FGVCx Cassava disease diagnosis

This is a new Fine Grained Visual Categorization-x (FGVCx) challenge for 2019 that will run as part of the FGVC6 workshop at CVPR

The Challenge

The goal of this task is to build a robust model that is able to distinguish between diseases in the Cassava plant. Cassava is an important food security crop for Africa grown by many small-holder farmers. The 4 diseases can be identified from the leaves of the plants. A possible extension of this challenge is to categorize the different severity levels of the diseases.

The key task with this dataset is to attempt a 5 class classification problem where you are attempting to distinguish between the 5 broad types of diseases; Cassava Mosaic Disease (CMD), Cassava Green Mite disease (CGM), Cassava Bacterial Blight (CBB), Cassava Brown Streak Disease (CBSD) and the health class (Figure above). In short, given a cassava leaf image can you predict the disease incidence.

A secondary task (to run after this challenge) would be to build a model that simultaneously predicts the disease incidence and severity. The data is labelled with the disease incidence and for each disease, the different types of example images of 5 types of severity scored 1-5; 1 being the class of a healthy leaf image, 5 being the class of a severely infected cassava plant as shown in the figure above.

Kaggle

For this challenge, we are using Kaggle to host the data and the leaderboard. Checkout the competition page here.

Dates (TBD)

Data Released	26-April-2019
Submission Server Open	26-April-2019
Submission Deadline	1-June-2019
Winners Announced	June 2019

Data

The data will consist of leaf images in each class. An abstracted reduced dataset of the images can be downloaded here.