Auto-Refining Reconstruction Algorithm for Recreation of Limited Angle Humanoid Depth Data.

Audrius Kulikajevas,Rytis Maskeliūnas,Marta Wlodarczyk-Sielicka,Robertas Damaševičius

doi:10.3390/s21113702

Audrius Kulikajevas, Rytis Maskeliūnas + Show 2 more

Open Access

https://doi.org/10.3390/s21113702

Copy DOI

Abstract

With the majority of research, in relation to 3D object reconstruction, focusing on single static synthetic object reconstruction, there is a need for a method capable of reconstructing morphing objects in dynamic scenes without external influence. However, such research requires a time-consuming creation of real world object ground truths. To solve this, we propose a novel three-staged deep adversarial neural network architecture capable of denoising and refining real-world depth sensor input for full human body posture reconstruction. The proposed network has achieved Earth Mover and Chamfer distances of and on synthetic datasets, respectively, which indicates on-par experimental results with other approaches, in addition to the ability of reconstructing from maskless real world depth frames. Additional visual inspection to the reconstructed pointclouds has shown that the suggested approach manages to deal with the majority of the real world depth sensor noise, with the exception of large deformities to the depth field.

Highlights

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations
With rapidly evolving field of threedimensional object reconstruction, such depth sensing systems [8,9] may be the key to boosting object reconstruction quality
We propose a novel unsupervised adversarial auto-refiner capable of full human body pointcloud reconstruction using only a single self-occluding depth view capable of reconstructing real depth sensor data with no masking nor any other direct interference

Summary

Introduction

Whereas regular cameras have already been widely adopted in various object detection tasks, depth sensors still have narrow range of research dedicated to them This can be attributed to them not being available for personal use until relatively recently with the introduction of the original Kinect sensor [1]. While the Kinect technology made the depth sensors affordable they have not had wide consumer adoption outside of entertainment [2,3], health-related applications were considered [4,5,6] This is, likely a rapidly shifting trend with more consumer grade sensors, such as Intel Realsense [7], being released and depth scanning systems being integrated as part of mobile devices. With rapidly evolving field of threedimensional object reconstruction, such depth sensing systems [8,9] may be the key to boosting object reconstruction quality

Methods

Results

Discussion

Conclusion