KUNGFU.AI guided Savimbo in developing a computer vision model to verify forest regrowth from ground-level images, laying the foundation for a farmer-centric carbon credit marketplace.
Savimbo envisioned a carbon credit marketplace like no other—one that directly rewarded farmers for preserving the Colombian Amazon. While most solutions relied on satellite imagery, Savimbo wanted something more ground-level and personal: a computer vision model trained on real forest images to verify regrowth and habitat health.
The mission seemed straightforward: train a classification model to differentiate between images of 1-year forest growth versus 5-year growth. But building this model from scratch required a strong foundation—data manipulation, model tuning, and pipeline development—all while ensuring accuracy and scalability. The end goal? Classify new forest images reliably and scale this solution for future rain forest regions.
Savimbo partnered with a third-party development team from Mexico to build and train the computer vision model. They brought in KUNGFU.AI as their technical advisor, ensuring the right machine learning approach and infrastructure were in place. While we didn’t write the code, we provided strategic guidance, helping them fine-tune their approach and avoid common pitfalls. Our collaboration ensured that the model was built efficiently and deployed smoothly.
Though still in the early stages, this initiative has the potential to be transformative. The model is a core component of Savimbo’s carbon credit marketplace, enabling them to verify the current state of forest regrowth and create a transparent, farmer-centric solution. Savimbo is already thinking about expanding the project to other rain forest regions, which would involve adapting the model to new vegetation types—proof that this solution could have an even broader impact in the fight against deforestation.
