Laser & Optoelectronics Progress, Volume. 59, Issue 18, 1815010(2022)
3D Reconstruction and Semantic Segmentation Method Combining PointNet and 3D-LMNet from Single Image
Figures & Tables(7)
Fig. 2. Schematic diagram of segmentation loss. (a) Forward loss; (b) backward loss
Fig. 3. Effect of different a value on CD, EMD and mIoU. (a) CD; (b) EMD; (c) mIoU
Fig. 4. Qualitative results on airplanes from ShapeNet. (a) Input image; (b) ground truth; (c) 3D-LMNet; (d) 3D-LMNet+PointNet; (e) our model
Fig. 5. Qualitative results on cars from ShapeNet. (a) Input image; (b) ground truth; (c) 3D-LMNet; (d) 3D-LMNet+PointNet; (e) our model
Fig. 6. Qualitative results on chairs from ShapeNet. (a) Input image; (b) ground truth; (c) 3D-LMNet; (d) 3D-LMNet+PointNet; (e) our model
Table 1. Experimental results on ShapeNet
View tableTable 1. Experimental results on ShapeNet
Category Metric 3D-LMNet+PointNet Our model Airplane CD 3.34 3.01 EMD 4.77 4.18 mIoU 62.98 71.12 Car CD 4.55 4.26 EMD 4.10 3.85 mIoU 60.01 61.68 Chair CD 6.41 5.91 EMD 8.02 7.83 mIoU 79.15 82.03 Mean CD 4.77 4.39 EMD 5.63 5.29 mIoU 67.38 71.61
Tools
Get Citation
Copy Citation Text
Hui Chen, Yong Tong, Li Zhu, Weibin Liang. 3D Reconstruction and Semantic Segmentation Method Combining PointNet and 3D-LMNet from Single Image[J]. Laser & Optoelectronics Progress, 2022, 59(18): 1815010
Paper Information
Category: Machine Vision
Received: Jun. 15, 2021
Accepted: Aug. 10, 2021
Published Online: Aug. 29, 2022
The Author Email: Chen Hui (chenhui@shiep.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20