UEVAVD: A Dataset for Developing UAV's Eye View Active Object Detection

TL;DR

A UAV's eye view active vision dataset named UEVAVD is released and improved the existing DRL-based AOD method by incorporating the inductive bias when learning the state representation and improving the existing DRL-based AOD method by incorporating the inductive bias when learning the state representation.

Abstract

Occlusion is a longstanding difficulty that challenges the UAV-based object detection. Many works address this problem by adapting the detection model. However, few of them exploit that the UAV could fundamentally improve detection performance by changing its viewpoint. Active Object Detection (AOD) offers an effective way to achieve this purpose. Through Deep Reinforcement Learning (DRL), AOD endows the UAV with the ability of autonomous path planning to search for the observation that is more conducive to target identification. Unfortunately, there exists no available dataset for developing the UAV AOD method. To fill this gap, we released a UAV's eye view active vision dataset named UEVAVD and hope it can facilitate research on the UAV AOD problem. Additionally, we improve the existing DRL-based AOD method by incorporating the inductive bias when learning the state representation. First, due to the partial observability, we use the gated recurrent unit to extract state representations from the observation sequence instead of the single-view observation. Second, we pre-decompose the scene with the Segment Anything Model (SAM) and filter out the irrelevant information with the derived masks. With these practices, the agent could learn an active viewing policy with better generalization capability. The effectiveness of our innovations is validated by the experiments on the UEVAVD dataset. Our dataset will soon be available at https://github.com/Leo000ooo/UEVAVD_dataset.

Figures (8)

• Figure 1: A demonstration of the AOD decision process by the UAV platform. The UAV's policy network gives moving instructions based on observations. It can autonomously decide whether to move and how to move to acquire an ideal observation while minimizing the movement cost.
• Figure 2: The five vehicle targets used in collecting the UEVAVD dataset.
• Figure 3: The sample distribution of target's location under three kinds of contexts.
• Figure 4: A demonstration of the predefined sampling points distribution and the UAV's moving directions. The lower airspace over the target is divided into discrete sections, and we record observations of the area of interest when the UAV is at these sampling points.
• Figure 5: A sample sequence illustrating how the observation and recognition result change with the view variation.