The article discusses the significance of omnidirectional 3D object detection for robots navigating outdoor environments. It highlights the development and evaluation of a system called Panopticus, designed for efficient detection in real-time. Extensive groundwork was laid with new datasets collected using a mobile 360° camera setup and tested across various edge devices. The paper also details the model's architecture, performance testing, robustness evaluation, and execution scheduling strategies aimed at adapting the system to diverse operational settings, ultimately supporting improved navigation capabilities in urban scenarios.
Omnidirectional 3D object detection is essential for outdoor applications where robots navigate environments like urban streets, enabling enhanced spatial awareness and interaction.
The evaluation process utilized various edge devices and real-world scenarios, showcasing the system's accuracy and runtime efficiency, crucial for practical implementation.
Our mobile testbed, equipped with a 360° camera and LiDAR, collected diverse datasets, enhancing the robustness of our model in real-world applications.
Through extensive performance testing on different datasets, we aimed to adapt our model for scalability and efficiency, focusing on the critical aspects of execution scheduling.
Collection
[
|
...
]