International Scholarly Research Notices / 2013 / Article / Tab 1

Research Article

Multimodal Markov Random Field for Image Reranking Based on Relevance Feedback

Table 1

MAP, precision at 5 (P5), 10 (P10), and 20 (P20) documents obtained by different parameter settings of the proposed multimodal MRF (columns 6–9), compared to the performance one would obtain with relevance feedback alone (columns 2–5). Column 1 indicates the parameter settings for the corresponding row as follows: the first value is , the number of relevance feedback images, is the value of for the visual energy function (i.e., ), and is the value of for the textual energy function (i.e., ).

Base IRS P5 P10 P20 MAP
SRI TIA-TXTIMG 0.4769 0.4538 0.3910 0.2359

Only relevance feedbackMultimodal MRF
Configuration P5P10 P20 MAP P5 P10 P20 MAP

; ; 0.5641 0.4897 0.4038 0.2486 0.6051 0.5436 0.4756 0.2852
; ; 0.7846 0.5872 0.4359 0.2742 0.7897 0.6282 0.5141 0.3123
; ; 1.0000 0.6795 0.4782 0.2984 1.0000 0.7256 0.5551 0.3344
; ; 1.0000 0.8692 0.5577 0.3352 1.0000 0.8846 0.6359 0.3752
; ; 1.0000 0.9744 0.6128 0.3560 1.0000 0.9744 0.6628 0.3919

Only relevance feedback Multimodal MRF
ConfigurationP5 P10 P20 MAP P5 P10 P20 MAP

; ; 0.5641 0.4897 0.4038 0.2486 0.6462 0.5744 0.5038 0.2960
; ; 0.7846 0.5872 0.4359 0.2742 0.8000 0.6282 0.5205 0.3153
; ; 1.0000 0.6795 0.4782 0.2984 1.0000 0.7256 0.5615 0.3375
; ; 1.0000 0.8692 0.5577 0.3352 1.0000 0.8718 0.5897 0.3470
; ; 1.0000 0.9744 0.6128 0.3560 1.0000 0.9744 0.6731 0.3936

Only relevance feedback Multimodal MRF
ConfigurationP5 P10 P20 MAP P5 P10 P20 MAP

; ; 0.5641 0.4897 0.4038 0.2486 0.6513 0.5821 0.5077 0.2956
; ; 0.7846 0.5872 0.4359 0.2742 0.8103 0.6487 0.5282 0.3154
; ; 1.0000 0.6795 0.4782 0.2984 1.0000 0.7256 0.5718 0.3358
; ; 1.0000 0.8692 0.5577 0.3352 1.0000 0.8923 0.6500 0.3801
; ; 1.0000 0.9744 0.6128 0.3560 1.0000 0.9744 0.6885 0.3966

Only relevance feedback Multimodal MRF
ConfigurationP5 P10 P20 MAP P5 P10 P20 MAP

; , 0.5641 0.4897 0.4038 0.2486 0.6718 0.5923 0.5167 0.3004
; , 0.7846 0.5872 0.4359 0.2742 0.8051 0.6436 0.5359 0.3142
; , 1.0000 0.6795 0.4782 0.2984 1.0000 0.7462 0.5987 0.3432
; , 1.0000 0.8692 0.5577 0.3352 1.0000 0.9000 0.6833 0.3851
; , 1.0000 0.9744 0.6128 0.3560 1.0000 0.9744 0.7218 0.4031

Only relevance feedback Multimodal MRF
ConfigurationP5 P10 P20 MAP P5 P10 P20 MAP

; , 0.5641 0.4897 0.4038 0.2486 0.6513 0.5282 0.4128 0.2569
; , 0.7846 0.5872 0.4359 0.2742 0.8410 0.6205 0.4423 0.2795
; , 1.0000 0.6795 0.4782 0.2984 1.0000 0.7000 0.4846 0.3016
; , 1.0000 0.8692 0.5577 0.3352 1.0000 0.8821 0.5654 0.3374
; , 1.0000 0.9744 0.6128 0.3560 1.0000 0.9744 0.6205 0.3580

We are committed to sharing findings related to COVID-19 as quickly and safely as possible. Any author submitting a COVID-19 paper should notify us at help@hindawi.com to ensure their research is fast-tracked and made available on a preprint server as soon as possible. We will be providing unlimited waivers of publication charges for accepted articles related to COVID-19.