A real-time flower pose estimation framework for computationally constrained robotic pollination systems.

Leverages the power of NeRFs and diffusion models for robust pose estimation in low textured environment.

Developed hybrid (handcrafted + learned) image point descriptors for lightweight detection and matching