Learning to Look Around: Intelligently Exploring Unseen Environments for Unknown Tasks

By Dinesh Jayaraman, Kristen Grauman. 2017

122 citations recorded by the source corpus.

It is common to implicitly assume access to intelligently captured inputs (e.g., photos from a human photographer), yet autonomously capturing good observations is itself a major challenge. We address the problem of learning to look around: if an agent has the ability to voluntarily acquire new views to observe its environment, how can it learn efficient exploratory behaviors to acquire informative visual observations? We propose a reinforcement learning solution, where the agent is rewarded for actions that reduce its uncertainty about the unobserved portions of its environment. Based on this principle, we develop a recurrent neural network-based approach to perform active completion of panoramic natural scenes and 3D object shapes. Crucially, the learned policies are not tied to any recognition task nor to the particular semantic content seen during training. As a result, 1) the learned "look around" behavior is relevant even for new tasks in unseen environments, and 2) training data acquisition involves no manual labeling. Through tests in diverse settings, we demonstrate that our approach learns useful generic policies that transfer to new unseen tasks and environments.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

A Survey of Embodied AI: From Simulators to Research Tasks (2021, background): “mas and CAD models [69, 70, 71, 72], which are usually curated datasets of human-taken photos [53].”
Learning Exploration Policies for Navigation (2019, background): “Jayaraman & Grauman (2018) use a related reward function (pixel reconstruction) to learn policies to look around (and subsequently solve tasks).”
ShapeCodes: Self-supervised Feature Learning by Lifting Views to Viewgrids (2017, methodology): “Our convolutional encoder-decoder [54] neural network architecture is similar to [18, 22, 25, 51].”
Occupancy Anticipation for Efficient Exploration and Navigation (2020, background): “More recent work uses deep learning to learn navigation [79,20,56,41,77,75,59,64] or exploration [48,6,57,28,51] policies in an end-to-end fashion.”
Aligning Cyber Space With Physical World: A Comprehensive Survey on Embodied AI (2024, unknown): “Current methods addressing active perception focus on interacting with the environment [86], [87] or by changing the viewing direction to obtain more visual information [88], [89], [90], [91].”
Semantic Curiosity for Active Visual Learning (2020, unknown): “to improve performance on the task by gaining more information about the environment. This has been instantiated in the context of object detection [1], amodal object detection [56], scene completion [29], and localization [11,22]. We consider the problem in a dierent setting and study how to ecient”
Sidekick Policy Learning for Active Visual Exploration (2018, methodology): “Following [30], we discretize the environment into a set of candidate viewpoints.”
An Exploration of Embodied Visual Exploration (2020, unknown): “In reconstruction [28, 42, 49], the agent is presented with uniformly spread query locations, and must predict the set of concepts present at each location (cf. Sec.”
Emergence of exploratory look-around behaviors through active observation completion (2019, methodology): “to look around nearby, yet during training it can access omnidirectional viewpoints. Existing methods facing this asymmetry simply restrict the agent to the same partial observability during training (8,10,23,25,26,29). In contrast, our sidekick approach introduces reward shaping and demonstrations ”