Optimal canal placement is critical to water resource management, especially in arid and semi-arid agricultural regions. In this paper, we present Hydro-Net, a new approach that utilizes high-resolution satellite images and advanced pathfinding methods to design optimal canals. By using the accurate segmentation of water bodies through deep learning models and comparing different pathfinding methods, Hydro-Net computes the optimal canal routes. The segmentation with deep learning reaches a validation accuracy of 78.5%. Hydro-Net’s pathfinding algorithm is superior to A* and Dijkstra, with a reduction in execution time of 30.8% and 47.5%, respectively, while still maintaining optimal path selection. Underpinning qualitative and quantitative analysis, Hydro-Net demonstrates the power of merging state-of-the-art segmentation with smart routing strategies. Being a data-driven and scalable solution, it not only boosts precision in water resource management but also facilitates efficient and sustainable irrigation planning, especially in water-scarce areas.