Modern agriculture faces enormous challenges associated with the growing demand for increased production, transformations in the political scenario, and changes in the environment. All of this puts pressure on producers who have no way to increase their arable land area. However, this challenging moment for farmer growth is a great opportunity to stand out in a market eager for more production. To face these new scenarios, farmers need to use their scarce resources efficiently, minimizing waste. The challenge becomes even more complicated for Brazilians, due to the complexity and constant changes in Brazilian legislation, deterioration of natural resources (soil and water in some regions), and constant invasions. Faced with this complex scenario and with various uncertainties, producers need new tools that assist in the automation of manual and routine functions, which despite being simple for a human were previously unthinkable to be performed by a computer. Even simple and repetitive activities like counting seedlings could not be performed automatically due to the inherent variation in the colors and shapes of the seedlings and the surrounding soil, combined with shadows from the canopies on the ground and images in poor conditions. Trees do not have one pattern, but several. All of these variations prevent writing a computer program that recognizes a tree in a satellite or drone image. Interestingly, a human easily recognizes a tree, but even the most advanced computers from a few years ago would have had difficulty identifying a tree in a photo. Why does a human have this ease in recognizing varied patterns and for a computer is it so difficult? Although it seems a somewhat obvious answer: humans learn. We learn from an early age what a tree is; we see hundreds of trees throughout our lives, in various sizes, shapes, colors, and textures. Therefore, when we see trees, situations, or similar objects, we associate them with an ideal object, meaning if it has a trunk, leaves, green or brown: it is a tree or looks like one. So how do we make a machine learn? In a similar way to humans, while humans learn from their experiences, machines learn from data or information. Currently, data abounds in agriculture. Thanks to the great revolution provided by precision agriculture with GPS-guided machines, sample collections, multispectral sensors, etc., today we have a lot of information that allows us to develop machines that can learn to perform from the simplest and most expensive tasks, like counting seedlings, to the most complex tasks, like determining the type and geographically quantifying the presence of pests. Despite the success that machine learning has achieved in medicine, agriculture has its peculiarities that make this sector a great challenge. Each region, soil, and crop is different. This increases the variation in the data; seedlings of different types of eucalyptus in different soils have much more variation than the same type of eucalyptus in one type of soil, making it difficult to develop a general solution from a single case. Machine learning can solve countless problems that farmers face today, but this is only possible if there is data available, lots of data, and preferably specific data.