ronakshah92 / iceberg-classifier Goto Github PK

The Iceberg classifier is a classification problem where the competition dataset contains the iceberg and the ship images data taken by satellite and numerical data for angle of projection taken for image. The objective is to accurately identify the iceberg from a ship. The dataset contains 3 input variables and a binary class variable: band1: This has 5625 elements, which are the flattened array for 75 x 75 pixels, as numbers and each number corresponds to polarization by HH channel(transmit/receive horizontally) band2: This also has same size of pixels as numbers and the number corresponds to polarization by HV channel (transmit horizontally and receive vertically). Inc angle: The angle represents the angle of projection from satellite to the object. Is iceberg: 0 for ship and 1 for iceberg. The prediction consists of the probability of every image in the test data of being an iceberg, which in turn quantifies the goodness of the model.