We introduce Grounded Situation Recognition (GSR), a task that requires producing structured semantic summaries of images describing: the primary activity, entities engaged in the activity with their roles (e.g. agent, tool), and bounding-box groundings of entities. GSR presents important technical challenges: identifying semantic saliency, categorizing and localizing a large and diverse set of entities, overcoming semantic sparsity, and disambiguating roles. Moreover, unlike in captioning, GSR is straightforward to evaluate. To study this new task we create the Situations With Groundings (SWiG) dataset which adds 278,336 bounding-box groundings to the 11,538 entity classes in the imsitu dataset. We propose a Joint Situation Localizer and find that jointly predicting situations and groundings with end-to-end training handily outperforms independent training on the entire grounding metric suite with relative gains between 8% and 32%. Finally, we show initial findings on three exciting future directions enabled by our models: conditional querying, visual chaining, and grounded semantic aware image retrieval. Code and data available at https://prior.allenai.org/projects/gsr.
When we humans look at a video of human-object interaction, we can not only infer what is happening but we can even extract actionable information and imitate those interactions. On the other hand, current recognition or geometric approaches lack the physicality of action representation. In this paper, we take a step towards a more physical understanding of actions. We address the problem of inferring contact points and the physical forces from videos of humans interacting with objects. One of the main challenges in tackling this problem is obtaining ground-truth labels for forces. We sidestep this problem by instead using a physics simulator for supervision. Specifically, we use a simulator to predict effects and enforce that estimated forces must lead to the same effect as depicted in the video. Our quantitative and qualitative results show that (a) we can predict meaningful forces from videos whose effects lead to accurate imitation of the motions observed, (b) by jointly optimizing for contact point and force prediction, we can improve the performance on both tasks in comparison to independent training, and (c) we can learn a representation from this model that generalizes to novel objects using few shot examples.
Recent single image unsupervised representation learning techniques show remarkable success on a variety of tasks. The basic principle in these works is instance discrimination: learning to differentiate between two augmented versions of the same image and a large batch of unrelated images. Networks learn to ignore the augmentation noise and extract semantically meaningful representations. Prior work uses artificial data augmentation techniques such as cropping, and color jitter which can only affect the image in superficial ways and are not aligned with how objects actually change e.g. occlusion, deformation, viewpoint change. In this paper, we argue that videos offer this natural augmentation for free. Videos can provide entirely new views of objects, show deformation, and even connect semantically similar but visually distinct concepts. We propose Video Noise Contrastive Estimation, a method for using unlabeled video to learn strong, transferable single image representations. We demonstrate improvements over recent unsupervised single image techniques, as well as over fully supervised ImageNet pretraining, across a variety of temporal and non-temporal tasks.
Sparsity in Deep Neural Networks (DNNs) is studied extensively with the focus of maximizing prediction accuracy given an overall parameter budget. Existing methods rely on uniform or heuristic non-uniform sparsity budgets which have sub-optimal layer-wise parameter allocation resulting in a) lower prediction accuracy or b) higher inference cost (FLOPs). This work proposes Soft Threshold Reparameterization (STR), a novel use of the soft-threshold operator on DNN weights. STR smoothly induces sparsity while learning pruning thresholds thereby obtaining a non-uniform sparsity budget. Our method achieves state-of-the-art accuracy for unstructured sparsity in CNNs (ResNet50 and MobileNetV1 on ImageNet-1K), and, additionally, learns non-uniform budgets that empirically reduce the FLOPs by up to 50%. Notably, STR boosts the accuracy over existing results by up to 10% in the ultra sparse (99%) regime and can also be used to induce low-rank (structured sparsity) in RNNs. In short, STR is a simple mechanism which learns effective sparsity budgets that contrast with popular heuristics.
Fine-tuning pretrained contextual word embedding models to supervised downstream tasks has become commonplace in natural language processing. This process, however, is often brittle: even with the same hyperparameter values, distinct random seeds can lead to substantially different results. To better understand this phenomenon, we experiment with four datasets from the GLUE benchmark, fine-tuning BERT hundreds of times on each while varying only the random seeds. We find substantial performance increases compared to previously reported results, and we quantify how the performance of the best-found model varies as a function of the number of fine-tuning trials. Further, we examine two factors influenced by the choice of random seed: weight initialization and training data order. We find that both contribute comparably to the variance of out-of-sample performance, and that some weight initializations perform well across all tasks explored. On small datasets, we observe that many fine-tuning trials diverge part of the way through training, and we offer best practices for practitioners to stop training less promising runs early. We publicly release all of our experimental data, including training and validation scores for 2,100 trials, to encourage further analysis of training dynamics during fine-tuning.
The ubiquity of embodied gameplay, observed in a wide variety of animal species including turtles and ravens, has led researchers to question what advantages play provides to the animals engaged in it. Mounting evidence suggests that play is critical in developing the neural flexibility for creative problem solving, socialization, and can improve the plasticity of the medial prefrontal cortex. Comparatively little is known regarding the impact of gameplay upon embodied artificial agents. While recent work has produced artificial agents proficient in abstract games, the environments these agents act within are far removed the real world and thus these agents provide little insight into the advantages of embodied play. Hiding games have arisen in multiple cultures and species, and provide a rich ground for studying the impact of embodied gameplay on representation learning in the context of perspective taking, secret keeping, and false belief understanding. Here we are the first to show that embodied adversarial reinforcement learning agents playing cache, a variant of hide-and-seek, in a high fidelity, interactive, environment, learn representations of their observations encoding information such as occlusion, object permanence, free space, and containment; on par with representations learnt by the most popular modern paradigm for visual representation learning which requires large datasets independently labeled for each new task. Our representations are enhanced by intent and memory, through interaction and play, moving closer to biologically motivated learning strategies. These results serve as a model for studying how facets of vision and perspective taking develop through play, provide an experimental framework for assessing what is learned by artificial agents, and suggest that representation learning should move from static datasets and towards experiential, interactive, learning.
In this paper we address the problem of visual reaction: the task of interacting with dynamic environments where the changes in the environment are not necessarily caused by the agents itself. Visual reaction entails predicting the future changes in a visual environment and planning accordingly. We study the problem of visual reaction in the context of playing catch with a drone in visually rich synthetic environments. This is a challenging problem since the agent is required to learn (1) how objects with different physical properties and shapes move, (2) what sequence of actions should be taken according to the prediction, (3) how to adjust the actions based on the visual feedback from the dynamic environment (e.g., when objects bouncing off a wall), and (4) how to reason and act with an unexpected state change in a timely manner. We propose a new dataset for this task, which includes 30K throws of 20 types of objects in different directions with different forces. Our results show that our model that integrates a forecaster with a planner outperforms a set of strong baselines that are based on tracking as well as pure model-based and model-free RL baselines.