Socially-intelligent agents are of growing interest in artificial intelligence. To this end, we need systems that can understand social relationships in diverse social contexts. Inferring the social context in a given visual scene not only involves recognizing objects, but also demands a more indepth understanding of the relationships and attributes of the people involved. To achieve this, one computational approach for representing human relationships and attributes is to use an explicit knowledge graph, which allows for high-level reasoning. We introduce a novel end-to-endtrainable neural network that is capable of generating a Social Relationship Graph -a structured, unified representation of social relationships and attributes -from a given input image. Our Social Relationship Graph Generation Network (SRG-GN) is the first to use memory cells like Gated Recurrent Units (GRUs) to iteratively update the social relationship states in a graph using scene and attribute context. The neural network exploits the recurrent connections among the GRUs to implement message passing between nodes and edges in the graph, and results in significant improvement over previous methods for social relationship recognition.

People naturally understand the emotions of-and often also empathize with-those around them. In this paper, we predict the emotional valence of an empathic listener over time as they listen to a speaker narrating a life story. We use the dataset provided by the OMG-Empathy Prediction Challenge, a workshop held in conjunction with IEEE FG 2019. We present a multimodal LSTM model with feature-level fusion and local attention that predicts empathic responses from audio, text, and visual features. Our best-performing model, which used only the audio and text features, achieved a concordance correlation coefficient (CCC) of .29 and .32 on the Validation set for the Generalized and Personalized track respectively, and achieved a CCC of .14 and .14 on the held-out Test set. We discuss the difficulties faced and the lessons learnt tackling this challenge.

Scene graph generation (SGG) aims to capture a wide variety of interactions between pairs of objects, which is essential for full scene understanding. Existing SGG methods trained on the entire set of relations fail to acquire complex reasoning about visual and textual correlations due to various biases in training data. Learning on trivial relations that indicate generic spatial configuration like 'on' instead of informative relations such as 'parked on' does not enforce this complex reasoning, harming generalization. To address this problem, we propose a novel framework for SGG training that exploits relation labels based on their informativeness. Our model-agnostic training procedure imputes missing informative relations for less informative samples in the training data and trains a SGG model on the imputed labels along with existing annotations. We show that this approach can successfully be used in conjunction with state-of-the-art SGG methods and improves their performance significantly in multiple metrics on the standard Visual Genome benchmark. Furthermore, we obtain considerable improvements for unseen triplets in a more challenging zero-shot setting.

No abstract

Acquiring accurate labels on large-scale datasets is both time consuming and expensive. To reduce the dependency of deep learning models on learning from clean labeled data, several recent research efforts are focused on learning with noisy labels. These methods typically fall into three design categories to learn a noise robust model: sample selection approaches, noise robust loss functions, or label correction methods. In this paper, we propose PARS: Pseudo-Label Aware Robust Sample Selection, a hybrid approach that combines the best from all three worlds in a joint-training framework to achieve robustness to noisy labels. Specifically, PARS exploits all training samples using both the raw/noisy labels and estimated/refurbished pseudo-labels via self-training, divides samples into an ambiguous and a noisy subset via loss analysis, and designs label-dependent noise-aware loss functions for both sets of filtered labels. Results show that PARS significantly outperforms the state of the art on extensive studies on the noisy CIFAR-10 and CIFAR-100 datasets, particularly on challenging high-noise and low-resource settings. In particular, PARS achieved an absolute 12% improvement in test accuracy on the CIFAR-100 dataset with 90% symmetric label noise, and an absolute 27% improvement in test accuracy when only 1/5 of the noisy labels are available during training as an additional restriction. On a real-world noisy dataset, Clothing1M, PARS achieves competitive results to the state of the art.