Active Object Detection for UAV Remote Sensing via Behavior Cloning and Enhanced Q-Network with Shallow Features
- Citation Author(s):
-
Zhuocheng ZouXikun HuPing Zhong
- Submitted by:
- Zhuocheng Zou
- Last updated:
- DOI:
- 10.21227/t9r0-s152
1033 views
- Categories:
- Keywords:
Abstract
CARLA-AOD is a novel aerial object detection dataset created using the CARLA autonomous driving simulator. It comprises 2,160 images from 18 diverse urban and rural scenarios, featuring 4 vehicle categories, 24 viewpoints, and 5 scales. This dataset aims to support research in aerial object detection, offering comprehensive viewpoint coverage and rich environmental diversity to enhance model performance and generalization.
Instructions:
The final CARLA-AOD dataset consists of 2,160 samples, with 66% used for training and 33% for testing. The labels are in CoCo format JSON annotations for 4 categories of vehicle targets in the images.
urban environment