Solving Visual Madlibs with Multiple Cues
This work addresses the challenge of visual question answering for researchers by improving accuracy on a specific dataset, though it is incremental as it builds on existing VQA methods with enhanced features and region selection.
The paper tackled the problem of answering fill-in-the-blank multiple choice questions on the Visual Madlibs dataset by using specialized image features from tasks like scene classification and person activity prediction, along with a method for selecting relevant image sub-regions, resulting in a significant improvement over the previous state of the art.
This paper focuses on answering fill-in-the-blank style multiple choice questions from the Visual Madlibs dataset. Previous approaches to Visual Question Answering (VQA) have mainly used generic image features from networks trained on the ImageNet dataset, despite the wide scope of questions. In contrast, our approach employs features derived from networks trained for specialized tasks of scene classification, person activity prediction, and person and object attribute prediction. We also present a method for selecting sub-regions of an image that are relevant for evaluating the appropriateness of a putative answer. Visual features are computed both from the whole image and from local regions, while sentences are mapped to a common space using a simple normalized canonical correlation analysis (CCA) model. Our results show a significant improvement over the previous state of the art, and indicate that answering different question types benefits from examining a variety of image cues and carefully choosing informative image sub-regions.