Antonio Loquercio
Antonio Loquercio
hi @mehmetgur, you only need annotations to validate the method. In case you only want to run it, you can use as input "masks" a uniform black image!
you could try to make the model smaller and train it again!
Hi! Thanks for your detailed feedback! The switching between modes could be a problem. One way to control it is decreasing [the acceptance threshold](https://github.com/uzh-rpg/agile_autonomy/blob/main/planner_learning/config/deploy_settings.yaml#L37). In this case more trajectories will...
You can use any data you want for validation. You could spit or re-collect new data.
the algorithm is by design an end-to-end controller. We use beta-flight as a low-level controller.
We use our in-house flight controller (https://github.com/uzh-rpg/rpg_quadrotor_control). However, you can also use the navigation policy also with PX4 or ardupilot, but you won't be able to fly as fast or...
We use our in-house quadrotor for experiments, and a TX2 for the flight stack and the path planner
we use Betaflight as a low-level controller. That is the same that is used by human-piloted drones.
We use a depth map, no explicit object detection
I was able to download it entirely. Does anybody else encounter this issue?