Oil Spill Classification

The dataset was meticulously developed using high-resolution satellite images of the ocean. Initially, these images were collected over various periods and regions, capturing a diverse array of oceanic conditions. Among these images, some depicted oil spills caused by either illegal or accidental dumping of oil into the ocean. On the other hand, others showed pristine ocean surfaces without any contamination.

To enhance the analysis and utility of these images, we divided each one into smaller sections or patches. This segmentation process enabled a more detailed examination of specific areas within the larger images. Then, we subjected these patches to advanced computer vision algorithms, which meticulously analyzed the visual data. The algorithms processed the images to extract a comprehensive vector of features, effectively summarizing the contents of each patch. These feature vectors encapsulated various aspects of the image patches, such as color, texture, shape, and other relevant visual attributes.

To distinguish between spill and non-spill using the features for a given ocean patch, it is essential to address two key issues:

Non-Spill: negative case, or majority class.
Oil Spill: positive case, or minority class.

This dataset is sourced from Kaggle.