Round 2
Problem statement for Round 2 is given below. Round 1 is closed now.
Use this form to submit your entry for Round 2 before 11:59 pm IST 31st January, 2019.
Problem statement: Implement a supervised classification algorithm to classify the given Sentinel-2 composite image for Valparai region into 8 distinct landcover classes.
The following datasets are provided to you in the starter script:
1. A Sentinel-2 composite image for the year 2018
2. Labeled training points for 8 landcover classes divided into training and validation fractions
Starter Script: code.earthengine.google.co.in/fd803aea07404af154bb9014fee9b120
Existing Approach:
We present a simple baseline implementation that uses the ALOS Global DSM along with available Sentinel-2 bands and implements a Random Forest classifier. The accuracy for this algorithm is at 81.285%.
Restrictions:
You cannot modify the training data in any way or add your own training data
You must use only publicly available data in the Earth Engine Data Catalog in your implementation
Judging Criteria:
Your submission will be judged on the following criteria
1. Classification accuracy as determined by classifying the validation points using your classifier
2. Creativity of the algorithm and dataset(s) used
3. Technical accuracy of code, EE functions used
Reference:
Check out the 'Supervised Classification' section of the Earth Engine User Guide: