Optical Instruments, Volume. 45, Issue 5, 62(2023)
A dual-branch guided network for depth completion
Figures & Tables(11)
Table 1. The comparison of metrics parameters for the input branch
View tableView in ArticleTable 1. The comparison of metrics parameters for the input branch
算法 RMSE↓ REL↓ MAE↓ δ （t= 1.05）↑ δ （t= 1.10）↑ δ （t=1.25）↑ Image-Guided 0.049 0.065 0.043 52.21 68.51 92.87 Mask-Guided 0.078 0.097 0.056 44.76 60.20 89.63 Joint-Guided 0.044 0.060 0.038 57.22 76.45 95.67
Table 2. The comparison of metrics parameters between different number of SRB
View tableView in ArticleTable 2. The comparison of metrics parameters between different number of SRB
N RMSE↓ REL↓ MAE↓ δ（t=1.05）↑ δ（t=1.10）↑ δ（t=1.25）↑ 3 0.063 0.095 0.058 49.21 67.12 93.11 4 0.050 0.073 0.042 56.35 75.31 95.27 5 0.042 0.057 0.035 60.48 81.20 96.83 6 0.041 0.057 0.035 60.71 81.32 96.80
Table 3. The comparison of metrics parameters with or without attention mechanism
View tableView in ArticleTable 3. The comparison of metrics parameters with or without attention mechanism
算法 RMSE↓ REL↓ MAE↓ δ （t=1.05）↑ δ （t=1.10）↑ δ （t=1.25）↑ JG 0.042 0.057 0.035 60.48 81.20 96.83 JG+Att 0.036 0.046 0.025 73.67 88.92 97.23 JG+Att- improved 0.032 0.045 0.024 74.35 89.71 97.90
Table 4. The comparison of metrics parameters on ClearGrasp dataset
View tableView in ArticleTable 4. The comparison of metrics parameters on ClearGrasp dataset
算法 RMSE↓ REL↓ MAE↓ δ （t=1.05）↑ δ （t=1.10）↑ δ （t=1.25）↑ JBF[ 26 ]0.389 0.530 0.358 27.61 37.28 51.32 MRF[ 27 ]0.347 0.497 0.311 38.35 52.63 65.19 AD[ 28 ]0.315 0.489 0.297 41.26 61.29 71.24 DenseDepth[ 30 ]0.270 0.428 0.259 18.67 34.34 58.29 DM[ 29 ]0.049 0.075 0.038 59.67 75.85 95.96 DeepCompletion[ 11 ]0.054 0.081 0.045 44.53 69.71 95.77 ClearGrasp[ 17 ]0.038 0.048 0.027 72.94 87.88 97.17 Ours 0.032 0.045 0.024 74.35 89.71 97.90
Table 5. The comparison of metrics parameters on TransCG dataset
View tableView in ArticleTable 5. The comparison of metrics parameters on TransCG dataset
算法 RMSE↓ REL↓ MAE↓ δ （t=1.05）↑ δ （t=1.10）↑ δ （t=1.25）↑ ClearGrasp[ 17 ]0.054 0.083 0.037 50.48 68.68 95.28 LIDF-Refine[ 18 ]0.019 0.034 0.015 78.22 94.26 99.80 DFNet[ 23 ]0.018 0.027 0.012 83.76 95.67 99.71 Ours 0.018 0.025 0.012 85.69 96.49 99.78
Table 6. The comparison of metrics parameters on cross-domain datasets
View tableView in ArticleTable 6. The comparison of metrics parameters on cross-domain datasets
训练/测试 算法 RMSE↓ REL↓ MAE↓ δ（t=1.05）↑ δ（t=1.10）↑ δ（t=1.25）↑ ClearGrasp/ TransCG ClearGrasp[ 17 ]0.061 0.108 0.049 33.59 54.73 92.48 LIDF-Refine[ 18 ]0.146 0.262 0.115 13.70 26.39 57.95 DFNet[ 23 ]0.048 0.088 0.039 38.65 62.42 95.28 Ours 0.036 0.070 0.034 45.47 75.21 96.02 TransCG/ ClearGrasp ClearGrasp[ 17 ]0.085 0.095 0.052 47.26 70.76 92.54 LIDF-Refine[ 18 ]0.152 0.225 0.139 9.86 20.63 46.02 DFNet[ 23 ]0.041 0.054 0.031 62.74 83.31 97.33 Ours 0.032 0.047 0.029 69.23 88.14 97.80
Tools
Get Citation
Copy Citation Text
Xiaofei QIN, Wenkai HU, Dongxian BAN, Hongyu GUO, Jing YU. A dual-branch guided network for depth completion[J]. Optical Instruments, 2023, 45(5): 62
Paper Information
Category:
Received: Dec. 17, 2022
Accepted: --
Published Online: Dec. 27, 2023
The Author Email:
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20