DualVD: An Adaptive Dual Encoding Model for Deep Visual Understanding in Visual Dialogue

Abstract

Different from Visual Question Answering task that requires to answer only one question about an image, Visual Dialogue involves multiple questions which cover a broad range of visual content that could be related to any objects, relationships or semantics. The key challenge in Visual Dialogue task is thus to learn a more comprehensive and semantic-rich image representation which may have adaptive attentions on the image for variant questions. In this research, we propose a novel model to depict an image from both visual and semantic perspectives. Specifically, the visual view helps capture the appearance-level information, including objects and their relationships, while the semantic view enables the agent to understand high-level visual semantics from the whole image to the local regions. Futhermore, on top of such multi-view image features, we propose a feature selection framework which is able to adaptively capture question-relevant information hierarchically in fine-grained level. The proposed method achieved state-of-the-art results on benchmark Visual Dialogue datasets. More importantly, we can tell which modality (visual or semantic) has more contribution in answering the current question by visualizing the gate values. It gives us insights in understanding of human cognition in Visual Dialogue.

Modeling Coreference Relations in Visual Dialog

“…We compare the results of our model with the results of the following previously published models obtained on the VisDial v1.0 dataset: LF (Das et al, 2017), HRE (Das et al, 2017), MN (Das et al, 2017), CorefNMN (Kottur et al, 2018), FGA (Schwartz et al, 2019), RVA (Niu et al, 2019), HA-CAN (Yang et al, 2019), Synergistic (Guo et al, 2019), DAN (Kang et al, 2019), Dual VD (Jiang et al, 2020), and CAG (Guo et al, 2020). To make a fair and transparent comparison, we do not compare our models with models which were pretrained on other vision-language datasets before finetuning them on the VisDial v1.0 dataset, all the more because the vision-language datasets used in the pretraining overlap with the testset of Visdial v1.0.…”

GoG: Relation-aware Graph-over-Graph Network for Visual Dialog

“…FGA (Schwartz et al, 2019) realizes a factor graph attention mechanism, which constructs the graph over all the multi-modal features and estimates their interactions. DualVD (Jiang et al, 2020b) constructs a scene graph to represent the image while embedding both relationships provided by (Zhang et al, 2019b) and original object detection features (Anderson et al, 2018). CAG (Guo et al, 2020) focuses on an iterative question-conditioned context-aware graph, including both fine-grained visual-objects and textualhistory semantics.…”

Exploring Contextual-Aware Representation and Linguistic-Diverse Expression for Visual Dialog

“…In this part, we compare the proposed method with some state-ofthe-art methods on VisDial V1.0 and V0.9: VisDial V1.0. First of all, we compare the performance of our approach with other state-of-the-art methods on VisDial v1.0 test set, including LF [3], HRE [3], MN [3], CorefMN [16], RvA [21], DL-61 [10], DVAN [7], DAN [14], VGNN [32], FGA [25], DualVD [13] and KBGN [12]. The reason that we do not compare our method with the BERT models is that they adopt large-scale multimodal data to pre-train, and we only use the VisDial training set to train the model.…”