Depth Estimation and Dense Reconstruction with the Monocular Camera (Tutor: Tianjun Zhang, wechat: z619850002)

Introduction

3D reconstruction based on the monocular vision is a classic task in the field of computer vision. In this project, you need to mainly focus on three tasks, including pose recovery, depth estimation and dense map construction.

                                                    Fig. 1: Depth Estimation

 

                                             Fig. 2: Dense Scene Reconstruction

 

   Fig. 3: Scene structure reconstruction incrementally using a single agent

Requirements

1)  Collecting at least one video sequence of images and recovering corresponding camera poses of these images based on any SFM or SLAM methods. After that, depth maps of key frames should be recovered by any proper algorithms. Finally, the dense map of the scene can be constructed incrementally.

2) When the video stream and corresponding poses are input, your system should be able to construct a dense map in real-time or quasi real-time, rather than offline. GPU is allowed.

Reference Materials