• Laser & Optoelectronics Progress
  • Vol. 59, Issue 18, 1810007 (2022)
Youjun Yue1, Jie Zhang1、*, Hui Zhao1、2, and Hongjun Wang1
Author Affiliations
  • 1Tianjin Key Laboratory of Control Theory & Applications in Complicated Systems, School of Electrical and Electronic Engineering, Tianjin University of Technology, Tianjin 300384, China
  • 2College of Engineering and Technology, Tianjin Agricultural University, Tianjin 300392, China
  • show less
    DOI: 10.3788/LOP202259.1810007 Cite this Article Set citation alerts
    Youjun Yue, Jie Zhang, Hui Zhao, Hongjun Wang. Real-Time Indoor Scene Layout Estimation Based on Improved Lightweight Network[J]. Laser & Optoelectronics Progress, 2022, 59(18): 1810007 Copy Citation Text show less
    Structure of the proposed network
    Fig. 1. Structure of the proposed network
    Applications of different parameters in ESA network
    Fig. 2. Applications of different parameters in ESA network
    Experimental results on LSUN dataset. (a) Original picture; (b) real layout; (c) layout obtained by proposed method
    Fig. 3. Experimental results on LSUN dataset. (a) Original picture; (b) real layout; (c) layout obtained by proposed method
    Experimental results on Hedau dataset. (a) Original picture; (b) real layout; (c) layout obtained by proposed method
    Fig. 4. Experimental results on Hedau dataset. (a) Original picture; (b) real layout; (c) layout obtained by proposed method
    Results of ablation experiment. (a) Original image; (b) real layout; (c) ablating layered monitoring (ALM); (d) ablating simple combination (ASC); (e) ablating semantic transfer (AST); (f) combination of three mechanisms (CTM)
    Fig. 5. Results of ablation experiment. (a) Original image; (b) real layout; (c) ablating layered monitoring (ALM); (d) ablating simple combination (ASC); (e) ablating semantic transfer (AST); (f) combination of three mechanisms (CTM)
    MethodPE /%End-to-endNetwork backbone
    Method in Ref.[524.23NoNo
    Method in Ref.[916.71NoFCN(VGG-16)
    Method in Ref.[1110.63NoFCN(VGG-16)
    Method in Ref.[129.31NoFCN(VGG-16)
    Method in Ref.[169.86YesEncode and decode
    Method in Ref.[1312.49NoEncode and decode

    Method in Ref.[19

    Method in Ref.[21

    7.79

    9.05

    Yes

    Yes

    SegNet×6

    Encode and decode

    Proposed method7.35YesESA(ResNet-34)
    Table 1. Performance evaluation of different methods on the LSUN dataset
    MethodPE /%
    Method in Ref.[521.20
    Method in Ref.[720.10
    Method in Ref.[812.80
    Method in Ref.[912.83
    Method in Ref.[119.73
    Method in Ref.[128.67
    Method in Ref.[168.36
    Method in Ref.[1312.70
    Method in Ref.[197.44
    Proposed method8.32
    Table 2. Performance evaluation of the different methods on the Hedau dataset
    MethodPE /%
    ALM7.84
    ASC7.67
    AST14.30
    CTM(ours)7.35
    Table 3. Results of the ablation study
    ad1d2d3PE /%
    0.250.250.250.258.81
    0.50.50.50.57.99
    0.70.70.70.78.07
    11117.72
    1.251.251.251.258.07
    10.50.50.57.67
    10.20.20.27.35
    10007.84
    Table 4. Results of different loss weights
    Youjun Yue, Jie Zhang, Hui Zhao, Hongjun Wang. Real-Time Indoor Scene Layout Estimation Based on Improved Lightweight Network[J]. Laser & Optoelectronics Progress, 2022, 59(18): 1810007
    Download Citation