Discovering Dynamic Salient Regions for Spatio-Temporal Graph Neural Networks
arXiv:2009.08427
Abstract
Graph Neural Networks are perfectly suited to capture latent interactions between various entities in the spatio-temporal domain (e.g. videos). However, when an explicit structure is not available, it is not obvious what atomic elements should be represented as nodes. Current works generally use pre-trained object detectors or fixed, predefined regions to extract graph nodes. Improving upon this, our proposed model learns nodes that dynamically attach to well-delimited salient regions, which are relevant for a higher-level task, without using any object-level supervision. Constructing these localized, adaptive nodes gives our model inductive bias towards object-centric representations and we show that it discovers regions that are well correlated with objects in the video. In extensive ablation studies and experiments on two challenging datasets, we show superior performance to previous graph neural networks models for video classification.
Accepted at Neural Information Processing Systems (NeurIPS 2021)
References in corpus (14)
- Neural Message Passing for Quantum Chemistry
- Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation
- Understanding the Effective Receptive Field in Deep Convolutional Neural Networks
- Interaction Networks for Learning about Objects, Relations and Physics
- Learning Deep Generative Models of Graphs
- MONet: Unsupervised Scene Decomposition and Representation
- What Can Neural Networks Reason About?
- Object Level Visual Reasoning in Videos
- Recurrent Independent Mechanisms
- On the Binding Problem in Artificial Neural Networks
- LatentGNN: Learning Efficient Non-local Relations for Visual Recognition
- Recurrent Space-time Graph Neural Networks
- S2RMs: Spatially Structured Recurrent Modules
- Better Set Representations For Relational Reasoning