Detecting Illegal Coffee Plantations and Deforestation
From mountainous jungles to low-lying dry forests, people are utilizing and accessing the forests of the Greater Mekong like never before. As recently as the early 70s, forests blanketed three quarters of the Greater Mekong. The last 30 years have seen more than 30% of that forest cover disappear.

Farmers are clearing forest cover for coffee plantations at an accelerating pace.

Forests landscapes and river systems in the Greater Mekong region are becoming increasingly fragmented. The remaining intact areas are increasingly smaller and isolated inside areas that have been heavily transformed.

This can make it difficult for some animals to find suitable habitat or to follow usual migration routes. It can also increase inbreeding, which in turn can increase the vulnerability of species to disease and other pressures.

When habitats are isolated, species also become more vulnerable to climate change, as their ability to move to areas with more favourable conditions becomes increasingly limited. -- WWF

Since illegal coffee plantations are often located deep in the forest, they are difficult to detect. Therefore, satellite images are now being used to detect these plantations. However, there are vast numbers of satellite images being collected and it is both expensive and inefficient to process these images using human experts.

Your job today is to build a machine learning model that can automatically detect the presence of a coffee plantation in a satellite image. You are provided with a small training set of labeled images (the label is 1 if there is a coffee plantation in the image, and the label is 0 if there is not).