3D Reconstruction and Accuracy Evaluation of Ancient Chinese Architectural Patches Based on Depth Learning from Single Image

Fig. 7. Mean value and standard deviation of the absolute depth error for the 15 typical scenes under the stereo image pair model and the motion estimation model

Download full size

Fig. 8. Depth maps and its error maps of the two models with large error

Download full size

Fig. 9. Depth maps and its error maps of the two models with small error

Download full size

Fig. 10. Mean value and standard deviation of the distance errors from the predicted point cloud to the ground truth after ICP registration optimization

Download full size

Fig. 11. Error distribution of point cloud(better result). (a) Stereo image pair model; (b) motion estimation model

Download full size

Fig. 12. Error distribution of point cloud(poor result). (a) Stereo image pair model; (b) motion estimation model

Download full size

Table 1. Used 7 different evaluation indexes
View table
Table 1. Used 7 different evaluation indexes
Evaluating index Definition
$A b s E$ $\frac{1}{|D|} \sum_{d \in D} |d^{'} - d|$
$A b s R e l$ $\frac{1}{|D|} \sum_{d \in D} |d^{'} - d| / d^{'}$
$R M S E$ $\sqrt[]{\frac{1}{|D|} \sum_{d \in D} {|d^{'} - d|}^{2}}$
$R M S E l o g$ $\sqrt[]{\frac{1}{|D|} \sum_{d \in D} {|l o g d^{'} - l o g d|}^{2}}$
$δ_{t}$ $\begin{array}{l} \frac{1}{|D|} |\{d \in D |m a x (\frac{d^{'}}{d}, \frac{d}{d^{'}}) < 1 . 25^{t}\}| \times 100 %, \\ (t = 1,2, 3) \end{array}$

Table 2. Comparison results of the stereo image pair model and the motion estimation model under the 7 indicators

View table

Table 2. Comparison results of the stereo image pair model and the motion estimation model under the 7 indicators

Scene	Parameter	AbsE		AbsRel		RMSElog		RMSE		δ₁		δ₂		δ₃
Scene	Parameter	S-DN	P-DN	S-DN	P-DN	S-DN	P-DN	S-DN	P-DN	S-DN	P-DN	S-DN	P-DN	S-DN	P-DN
S1	Mean	92.059	69.396	0.044	0.033	0.019	0.015	130.047	101.968	0.992	0.996	0.998	0.999	0.999	0.999
S1	Std	20.119	13.009	0.011	0.008	0.004	0.003	30.701	26.728	0.012	0.007	0.005	0.003	0.002	0.002
S2	Mean	142.004	85.449	0.069	0.042	0.031	0.018	218.779	134.206	0.939	0.986	0.986	0.997	0.998	1.000
S2	Std	85.910	50.506	0.033	0.022	0.017	0.010	104.946	56.033	0.086	0.017	0.021	0.004	0.005	0.001
S3	Mean	76.176	47.532	0.143	0.062	0.044	0.026	172.549	116.678	0.890	0.970	0.967	0.987	0.986	0.991
S3	Std	47.523	33.026	0.104	0.036	0.027	0.015	134.621	79.858	0.107	0.075	0.036	0.02	0.018	0.012
S4	Mean	81.869	53.993	0.053	0.037	0.023	0.016	118.610	84.116	0.985	0.991	0.995	0.999	0.998	0.999
S4	Std	27.412	18.137	0.018	0.013	0.008	0.005	46.794	31.098	0.022	0.012	0.011	0.003	0.005	0.001
S5	Mean	79.434	53.465	0.048	0.035	0.021	0.015	127.196	87.650	0.986	0.990	0.995	0.998	0.999	0.999
S5	Std	13.332	15.185	0.007	0.011	0.003	0.005	28.824	31.470	0.012	0.009	0.005	0.003	0.002	0.001
S6	Mean	73.650	51.223	0.060	0.044	0.024	0.018	146.474	121.922	0.977	0.981	0.986	0.988	0.993	0.995
S6	Std	32.010	24.641	0.034	0.026	0.009	0.008	102.214	92.644	0.021	0.018	0.016	0.013	0.012	0.009
S7	Mean	147.347	77.696	0.119	0.061	0.040	0.023	262.657	153.315	0.941	0.969	0.960	0.984	0.968	0.990
S7	Std	77.375	39.067	0.099	0.046	0.024	0.014	199.390	110.325	0.073	0.039	0.062	0.025	0.052	0.017
S8	Mean	61.501	45.311	0.059	0.039	0.024	0.017	112.967	91.378	0.965	0.987	0.991	0.994	0.995	0.997
S8	Std	29.071	15.768	0.036	0.016	0.013	0.007	52.998	42.383	0.048	0.019	0.015	0.012	0.009	0.008
S9	Mean	61.198	57.930	0.036	0.034	0.016	0.015	89.288	82.843	0.991	0.995	0.999	0.999	1.000	1.000
S9	Std	17.013	13.090	0.011	0.007	0.004	0.003	25.556	19.684	0.013	0.006	0.003	0.002	0.001	0.000
S10	Mean	98.554	79.701	0.057	0.046	0.024	0.020	148.852	118.570	0.979	0.988	0.995	0.999	0.999	1.000
S10	Std	28.063	27.118	0.019	0.016	0.008	0.007	39.469	35.626	0.024	0.014	0.006	0.002	0.002	0.001
S11	Mean	112.137	94.995	0.053	0.045	0.022	0.020	171.664	149.130	0.984	0.986	0.997	0.998	0.999	1.000
S11	Std	16.456	21.645	0.008	0.009	0.003	0.004	30.734	31.393	0.011	0.010	0.006	0.003	0.002	0.001
S12	Mean	60.529	46.035	0.065	0.044	0.024	0.018	111.509	89.939	0.967	0.981	0.988	0.993	0.992	0.995
S12	Std	67.468	39.070	0.155	0.057	0.030	0.016	102.026	68.838	0.069	0.050	0.048	0.031	0.040	0.026
S13	Mean	67.791	51.239	0.043	0.034	0.019	0.015	109.101	89.085	0.989	0.992	0.996	0.998	0.998	0.999
S13	Std	21.200	14.242	0.013	0.010	0.005	0.004	37.649	34.578	0.011	0.009	0.007	0.004	0.006	0.003
S14	Mean	138.875	62.554	0.056	0.026	0.025	0.011	161.117	83.849	0.996	0.999	0.999	1.000	1.000	1.000
S14	Std	49.425	10.958	0.017	0.004	0.008	0.002	47.469	13.856	0.009	0.002	0.002	0.000	0.001	0.000
S15	Mean	37.833	31.337	0.044	0.035	0.017	0.015	76.269	58.719	0.985	0.988	0.992	0.994	0.996	0.998
S15	Std	20.797	16.653	0.033	0.022	0.009	0.008	64.297	38.356	0.026	0.024	0.021	0.015	0.013	0.007

Table 3. Mean value and standard deviation of the distance errors from the predicted point cloud to the ground truth under the 15 typical scenes

View table

Table 3. Mean value and standard deviation of the distance errors from the predicted point cloud to the ground truth under the 15 typical scenes

Scene	Mean-ICP P-DN	Std-ICP P-DN	Mean-ICP S-DN	Std-ICP- S-DN
S1	52.8	22.1	60.8	15.7
S2	110.2	30.1	112.2	20.0
S3	40.0	13.8	52.8	39.9
S4	49.2	15.9	47.4	15.4
S5	50.9	26.1	51.9	33.1
S6	60.1	37.9	59.4	37.0
S7	49.6	17.1	52.4	14.5
S8	33.4	21.0	41.92	23.5
S9	40.4	9.8	53.7	13.1
S10	150.8	14.5	154.2	21.0
S11	134.0	22.9	128.8	26.1
S12	54.3	21.4	54.9	22.5
S13	47.0	14.3	44.67	12.0
S14	54.9	12.3	55.7	12.3
S15	27.0	16.7	29.3	16.0

Tools

Get Citation

Copy Citation Text

Lihua Hu, Wenzhuang Yin, Siyuan Xing, Jifu Zhang, Qiulei Dong, Zhanyi Hu. 3D Reconstruction and Accuracy Evaluation of Ancient Chinese Architectural Patches Based on Depth Learning from Single Image[J]. Laser & Optoelectronics Progress, 2022, 59(14): 1415020

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Machine Vision

Received: Mar. 8, 2022

Accepted: Mar. 29, 2022

Published Online: Jul. 1, 2022

The Author Email: Zhanyi Hu (huzy@nlpr.ia.ac.cn)

DOI:10.3788/LOP202259.1415020

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Used 7 different evaluation indexes

Table 1. Used 7 different evaluation indexes

Table 2. Comparison results of the stereo image pair model and the motion estimation model under the 7 indicators

Table 2. Comparison results of the stereo image pair model and the motion estimation model under the 7 indicators

Table 3. Mean value and standard deviation of the distance errors from the predicted point cloud to the ground truth under the 15 typical scenes

Table 3. Mean value and standard deviation of the distance errors from the predicted point cloud to the ground truth under the 15 typical scenes