Towards Efficient Multiview Object Detection with Adaptive Action Prediction

Active vision is a desirable perceptual feature for robots. Existing approaches usually make strong assumptions about the task and environment, thus are less robust and efficient. This study proposes an adaptive view planning approach to boost the efficiency and robustness of active object detection. We formulate the multi-object detection task as an active multiview object detection problem given the initial location of the objects. Next, we propose a novel adaptive action prediction (A2P) method built on a deep Q-learning network with a dueling architecture. The A2P method is able to perform view planning based on visual information of multiple objects; and adjust action ranges according to the task status. Evaluated on the AVD dataset, A2P leads to 21. 9% increase in detection accuracy in unfamiliar environments, while improving efficiency by 22. 7%. On the T-LESS dataset, multi-object detection boosts efficiency by more than 30% while achieving equivalent detection accuracy.

Authors

• Qianli Xu
• Fen Fang
• Nicolas Gauthier
• Wenyu Liang
• Yan Wu 0002
• Liyuan Li
• Joo Hwee Lim

Keywords

• Visualization
• Automation
• Adaptive systems
• Conferences
• Object detection
• Prediction methods
• Robustness
• Activity Prediction
• Adaptation Actions
• Detection Accuracy
• Multiple Objects
• Object Location
• Objective Information
• Deep Q-learning
• Convolutional Neural Network
• Path Length
• Convolutional Layers
• Search Space
• Intersection Over Union
• Object Recognition
• Bounding Box
• Single Object
• Type Of Action
• Reward Function
• Markov Decision Process
• Object Tracking
• Unseen Environments
• Objects In The Scene
• Multi-step Method
• Multi-step Prediction
• Object Bounding Boxes
• Average Path Length
• Policy Learning
• Benchmark Methods
• Intermediate State
• Individual Objects