| Popis: |
Unmanned Aerial Vehicle is an aircraft that operates and flies without a human pilot. It can reach at places where humans may not reach easily, such as search and rescue operations, earthquake mapping and flood mapping. It is additionally valuable for autonomous tasks such as the delivery of any item and target tracking which requires self-governing navigation. Motivated by the mentioned applications, in this paper we present a deep learning model for self-governing navigation of UAV. Our model exploits transfer learning from a well-known network architecture called MobileNet and it is trained on a dataset of images, collected from the various indoor environments. From an image, the model classifies actions such as either to go forward or to stop. Furthermore, after some experiments and results, we infer that among all Convolution Neural Network (CNN) architectures, the MobileNet architecture is ideal and appropriate for our purposed approach. |
