ConsNet: Learning Consistency Graph for Zero-Shot Human-Object Interaction Detection
This work improves HOI detection for computer vision applications by handling rare or unseen interactions, though it is incremental as it builds on existing graph-based and embedding methods.
The paper tackles the problem of Human-Object Interaction (HOI) Detection by addressing long-tail distribution and polysemy of action labels, proposing ConsNet, a knowledge-aware framework that uses a consistency graph and Graph Attention Networks to propagate knowledge, and it outperforms state-of-the-art methods on V-COCO and HICO-DET datasets under fully-supervised and zero-shot settings.
We consider the problem of Human-Object Interaction (HOI) Detection, which aims to locate and recognize HOI instances in the form of <human, action, object> in images. Most existing works treat HOIs as individual interaction categories, thus can not handle the problem of long-tail distribution and polysemy of action labels. We argue that multi-level consistencies among objects, actions and interactions are strong cues for generating semantic representations of rare or previously unseen HOIs. Leveraging the compositional and relational peculiarities of HOI labels, we propose ConsNet, a knowledge-aware framework that explicitly encodes the relations among objects, actions and interactions into an undirected graph called consistency graph, and exploits Graph Attention Networks (GATs) to propagate knowledge among HOI categories as well as their constituents. Our model takes visual features of candidate human-object pairs and word embeddings of HOI labels as inputs, maps them into visual-semantic joint embedding space and obtains detection results by measuring their similarities. We extensively evaluate our model on the challenging V-COCO and HICO-DET datasets, and results validate that our approach outperforms state-of-the-arts under both fully-supervised and zero-shot settings. Code is available at https://github.com/yeliudev/ConsNet.