In this session we'll share the history, victories and bruises from this multi-year quest of using AI to improve the vegetation maintenance planning process for the low voltage grid. E-Redes, the main distributor of electrical power in Portugal, and NTT Data teamed up to improve the process of identifying and managing the vegetation that threatens the power grid. We'll do a deep dive into the neural network architectures, bayesian processes, handcrafted heuristics and also cover the project-changing benefits of VQA in the labelling and evaluation pipelines. We'll conclude by sharing key takeaways and best practices, offering valuable insights for those seeking to succeed in their own computer vision projects.