DMCA
Annotator rationales for visual recognition (2011)
| Venue: | In ICCV |
| Citations: | 16 - 3 self |
BibTeX
@INPROCEEDINGS{Donahue11annotatorrationales,
author = {Jeff Donahue and Kristen Grauman},
title = {Annotator rationales for visual recognition},
booktitle = {In ICCV},
year = {2011},
pages = {1395--1402},
publisher = {IEEE}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Traditional supervised visual learning simply asks annotators “what ” label an image should have. We propose an approach for image classification problems requiring subjective judgment that also asks “why”, and uses that information to enrich the learned model. We develop two forms of visual annotator rationales: in the first, the annotator highlights the spatial region of interest he found most influential to the label selected, and in the second, he comments on the visual attributes that were most important. For either case, we show how to map the response to synthetic contrast examples, and then exploit an existing large-margin learning technique to refine the decision boundary accordingly. Results on multiple scene categorization and human attractiveness tasks show the promise of our approach, which can more accurately learn complex categories with the explanations behind the label choices. Is this anchor doing a Is this scene from a Is this woman serious or light story? comedy or a drama? “hot or not?” Are these TV characters Is this scene a Is this figure friends or nemeses? lounge or a bedroom? skater’s form good? Figure 1. Main Premise: Subjective or complex image classification tasks such as those depicted above may require deeper insight from human annotators than the usual category labels. We propose to ask for spatial or attribute-based rationales for the labels chosen, and augment a large-margin classifier objective to exploit both the labels and these explanations. 1.
Keyphrases
visual recognition annotator rationale human annotator tv character visual attribute large-margin classifier objective light story skater form image classification problem spatial region learn complex category human attractiveness task subjective judgment large-margin learning technique attribute-based rationale synthetic contrast example visual annotator rationale usual category label main premise visual learning complex image classification task multiple scene categorization learned model label choice
