In this paper, we propose a new pipeline of training a monocular UAV to fly a collision-free trajectory along the dense forest trail. As gathering high-precision images in the real world is expensive and the off-the-shelf dataset has some deficiencies, we collect a new dense forest trail dataset in a variety of simulated environment in Unreal Engine. Then we formulate visual perception of forests as a classification problem. A ResNet-18 model is trained to decide the moving direction frame by frame. To transfer the learned strategy to the real world, we construct a ResNet-18 adaptation model via multi-kernel maximum mean discrepancies to leverage the relevant labelled data and alleviate the discrepancy between simulated and real environment. Simulation and real-world flight with a variety of appearance and environment changes are both tested. The ResNet-18 adaptation and its variant model achieve the best result of 84.08% accuracy in reality.
Full paper can be available at here: 《Learning Transferable UAV for Forest Visual Perception》
Folders trail1 and trail2 contain the dataset retrieved from simulation. Folders 001..005 contain data gathered in real world for testing the system. Here are some samples can be download for preview.
Dataset is available as a single archive or as separated archives:
We applied the trained adaptation model to a UAV in the simulated world. Here we provide two representative videos. In the first one, the UAV completed the flight in the trail1 without any collision. In the second one, we presented some failure clips of four seasons in which the UAV do not finish the trail. All videos use 5x speed.
In two videos, we follow the perspective of the UAV in the main window. Also, the first-person view of the onboard camera is provided in the sub-window. Note that, since the UAV decides the moving direction frame by frame, there is some bumpy between each step.
If you have some questions, or you want to access the source code, please feel free to contact me.