Laser & Optoelectronics Progress, Volume. 58, Issue 14, 1410010(2021)
Colorful 3D Reconstruction from Single Image Based on Deep Learning
Figures & Tables(11)
Fig. 6. Comparison of predictions of different 3D models. (a) Original images; (b) 3D-R2N2; (c) PSGN; (d) OccNet; (e) SoftRas_O; (f) designed model
Fig. 7. Reconstruction results of different color 3D models. (a1) (a2) Original images; (b1) (b2) results of reconstruction
Table 1. Comparison of indicators of different attention mechanism
View tableTable 1. Comparison of indicators of different attention mechanism
Attention mechanism IOU SSIM MSE SoftRas_O 0.576 0.548 0.175 CBAM_S 0.589 0.579 0.161 SENet 0.593 0.563 0.186 CBAM 0.593 0.578 0.163
Table 2. Comparison of IOU indicators of different CBAMs and SoftRas_O
View tableTable 2. Comparison of IOU indicators of different CBAMs and SoftRas_O
Category IOU SoftRas_O CBAM 1-2 CBAM 3-4 CBAM 1-4 Airplane 0.603 0.646 0.648 0.648 Bench 0.444 0.461 0.456 0.454 Cabinet 0.616 0.629 0.613 0.633 Car 0.672 0.665 0.669 0.675 Chair 0.481 0.492 0.478 0.493 Display 0.556 0.571 0.554 0.563 Lamp 0.444 0.463 0.460 0.459 Loudspeaker 0.615 0.631 0.635 0.636 Rifle 0.643 0.677 0.683 0.674 Sofa 0.619 0.618 0.628 0.630 Table 0.453 0.487 0.476 0.470 Telephone 0.741 0.744 0.776 0.766 Vessel 0.589 0.611 0.610 0.612 Mean 0.575508 0.59192 0.59123 0.59330
Table 3. Comparison of IOU indicators of different models
View tableTable 3. Comparison of IOU indicators of different models
Category IOU 3D-R2N2 OccNet SoftRas_O Our model Airplane 0.426 0.547 0.603 0.648 Bench 0.373 0.452 0.444 0.454 Cabinet 0.676 0.732 0.616 0.633 Car 0.661 0.731 0.672 0.675 Chair 0.439 0.502 0.481 0.493 Display 0.440 0.479 0.556 0.563 Lamp 0.281 0.370 0.444 0.459 Loudspeaker 0.611 0.653 0.615 0.636 Rifle 0.375 0.458 0.643 0.674 Sofa 0.626 0.671 0.619 0.630 Table 0.420 0.506 0.453 0.470 Telephone 0.611 0.709 0.741 0.766 Vessel 0.482 0.521 0.589 0.612 Mean 0.493 0.564 0.571 0.593
Table 4. Comparison results of SSIM and MSE between designed model and SoftRas_O
View tableTable 4. Comparison results of SSIM and MSE between designed model and SoftRas_O
Category SSIM MSE SoftRas_O Our model SoftRas_O Our model Airplane 0.627 0.610 0.084 0.090 Bench 0.475 0.493 0.186 0.183 Cabinet 0.565 0.587 0.188 0.182 Car 0.584 0.592 0.197 0.184 Chair 0.442 0.486 0.240 0.222 Display 0.453 0.488 0.264 0.250 Lamp 0.537 0.749 0.156 0.068 Loudspeaker 0.569 0.591 0.218 0.219 Rifle 0.592 0.592 0.081 0.070 Sofa 0.489 0.494 0.230 0.228 Table 0.537 0.548 0.198 0.197 Telephone 0.654 0.673 0.136 0.133 Vessel 0.600 0.617 0.096 0.098 Mean 0.548 0.578 0.175 0.163
Tools
Get Citation
Copy Citation Text
Yuzheng Zhu, Yaping Zhang, Qiaosheng Feng. Colorful 3D Reconstruction from Single Image Based on Deep Learning[J]. Laser & Optoelectronics Progress, 2021, 58(14): 1410010
Paper Information
Category: Image Processing
Received: Oct. 10, 2020
Accepted: Nov. 14, 2020
Published Online: Jul. 6, 2021
The Author Email: Yaping Zhang (zhangyp@ynnu.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20