TY - GEN
T1 - QuadSampling
T2 - 18th International Conference on Computer-Aided Design and Computer Graphics, CAD/Graphics 2023
AU - Hu, Xu Qiang
AU - Wang, Yu Ping
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.
PY - 2024
Y1 - 2024
N2 - Implicit neural representations have shown potential advantages in 3D reconstruction. But implicit neural 3D reconstruction methods require high-performance graphical computing power, which limits their application on low power consumption platforms. Remote 3D reconstruction framework can be employed to address this issue, but the sampling method needs to be further improved. We present a novel sampling method, QuadSampling, for remote implicit neural 3D reconstruction. By hierarchically sampling pixels within blocks with larger loss value, QuadSampling can result in larger average loss and help the neural learning process by better representing the shape of regions with different loss value. Thus, under the same amount of transmission, our QuadSampling can obtain more accurate and complete implicit neural representation of the scene. Extensive evaluations show that comparing with prior methods (i.e. random sampling and active sampling), our QuadSampling framework can improve the accuracy by up to 4%, and the completion ratio by about 1–2%.
AB - Implicit neural representations have shown potential advantages in 3D reconstruction. But implicit neural 3D reconstruction methods require high-performance graphical computing power, which limits their application on low power consumption platforms. Remote 3D reconstruction framework can be employed to address this issue, but the sampling method needs to be further improved. We present a novel sampling method, QuadSampling, for remote implicit neural 3D reconstruction. By hierarchically sampling pixels within blocks with larger loss value, QuadSampling can result in larger average loss and help the neural learning process by better representing the shape of regions with different loss value. Thus, under the same amount of transmission, our QuadSampling can obtain more accurate and complete implicit neural representation of the scene. Extensive evaluations show that comparing with prior methods (i.e. random sampling and active sampling), our QuadSampling framework can improve the accuracy by up to 4%, and the completion ratio by about 1–2%.
KW - Implicit Neural Representation
KW - Remote 3D Reconstruction
KW - Sampling Method
UR - http://www.scopus.com/inward/record.url?scp=85185845579&partnerID=8YFLogxK
U2 - 10.1007/978-981-99-9666-7_21
DO - 10.1007/978-981-99-9666-7_21
M3 - Conference contribution
AN - SCOPUS:85185845579
SN - 9789819996650
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 314
EP - 328
BT - Computer-Aided Design and Computer Graphics - 18th International Conference, CAD/Graphics 2023, Proceedings
A2 - Hu, Shi-Min
A2 - Cai, Yiyu
A2 - Rosin, Paul
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 19 August 2023 through 21 August 2023
ER -