Pose-Assisted Multi-Camera Collaboration System[1] is a novel method, which enables a camera to cooperate with the others by sharing camera poses for AOT(active object tracking). There are some details of reading and implementing it.

Contents

• Paper & Code & note
• Paper understanding
• Code understanding
• Note
• References

Paper & Code & note

Paper: Pose-Assisted Multi-Camera Collaboration for Active Object Tracking(AAAI 2020 paper)
Code: [Pytorch][Updating]
Note: PAMCC-AOT

Paper

Abstract

1. PAMCC-AOT is proposed to solve complex scenarios problems.
2. The vision-based controller tracks targets based on observed images.
3. The pose-based controller moves the camera in accordance to the poses of the other cameras.
4. At each step, the switcher decides which action to take from the two controllers according to the visibility of the target.

Problem Description

AOT: a tracker is able to control its motion so as to follow a target autonomously.
Problems: high complexity of environments and limitation of camera mobility.

Problem Solution

1. It extend the independent AOT to the CMC-AOT.
2. They proposed PAMCC-AOT sharing camera poses.
3. They provided a set of 3D environments.

Conceptual Understanding

In the system, each camera is equipped with two controllers and a switcher.

1. Vision-based Controller: it serves as an image processor and guides the camera to execute policy based on image observation.
2. Pose-based Controller: it helps the camera who receives an imperfect observation to execute policy based on the supplementary pose information provided by other cameras.
3. Switcher: it makes the camera switch between the vision-based controller and pose-based controller properly.

Experiments

Code

[Updating]

Note

References

[1] Li, Jing, et al. “Pose-Assisted Multi-Camera Collaboration for Active Object Tracking.” arXiv preprint arXiv:2001.05161 (2020).