Recommendation for the Best Object Detection Model for BeagleBone AI-64 (TDA4VM) for Use in Drones

I am currently working on an object detection project using the BeagleBone AI-64, which is based on the TDA4VM SoC. The goal is to deploy this system on drones for real-time detection of objects such as buildings, vehicles, and infrastructure.

Could anyone kindly recommend the best-suited object detection model that performs well on this platform, considering the constraints of embedded systems and the need for real-time performance?

Additionally, any guidance on model optimization (e.g., quantization, pruning) or deployment strategies for TDA4VM would be greatly appreciated.