Reinforcement Learning of Active Vision for Manipulating Objects under Occlusions

Cheng, Ricson; Agarwal, Amit; Fragkiadaki, Katerina

doi:10.48550/arxiv.1811.08067

Cited by 6 publications

(8 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [34] an active vision dataset and a reinforcement learning based baseline to explore the environment to detect objects are proposed. [35] presents a method to jointly learn a policy for both grasping and viewing. The method uses a simulated environment which detects objects depending on the occlusion rate.…”

Section: Related Workmentioning

confidence: 99%

Active 6D Multi-Object Pose Estimation in Cluttered Scenarios with Deep Reinforcement Learning

Sock

Garcia-Hernando

Kim

2020

2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

View full text Add to dashboard Cite

In this work, we explore how a strategic selection of camera movements can facilitate the task of 6D multi-object pose estimation in cluttered scenarios while respecting real-world constraints important in robotics and augmented reality applications, such as time and distance travelled. In the proposed framework, a set of multiple object hypotheses is given to an agent, which is inferred by an object pose estimator and subsequently spatio-temporally selected by a fusion function that makes use of a verification score that circumvents the need of ground-truth annotations. The agent reasons about these hypotheses, directing its attention to the object which it is most uncertain about, moving the camera towards such an object. Unlike previous works that propose short-sighted policies, our agent is trained in simulated scenarios using reinforcement learning, attempting to learn the camera moves that produce the most accurate object poses hypotheses for a given temporal and spatial budget, without the need of viewpoints rendering during inference. Our experiments show that the proposed approach successfully estimates the 6D object pose of a stack of objects in both challenging cluttered synthetic and real scenarios, showing superior performance compared to strong baselines.

show abstract

Section: Related Workmentioning

confidence: 99%

Active 6D Multi-Object Pose Estimation in Cluttered Scenarios with Deep Reinforcement Learning

Sock

Garcia-Hernando

Kim

2020

2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

View full text Add to dashboard Cite

show abstract

“…We do not plan based on detailed geometric models, but instead leverage coarser geometric information, together with semantic information, and between hierarchical levels of the 3D scene, to guide the search. Recent works used neural networks to visually reason about piles of objects and learn how to manipulate it with the objective of finding a target object [1,8,[23][24][25]. Different to ours, their method is restricted to the manipulation of the pile and does not leverage semantic cues.…”

Section: Related Workmentioning

confidence: 99%

Semantic and Geometric Modeling with Neural Message Passing in 3D Scene Graphs for Hierarchical Mechanical Search

Kurenkov¹,

Roberto²,

Ichnowski³

et al. 2020

Preprint

View full text Add to dashboard Cite

Searching for objects in indoor organized environments such as homes or offices is part of our everyday activities. When looking for a target object, we jointly reason about the rooms and containers the object is likely to be in; the same type of container will have a different probability of having the target depending on the room it is in. We also combine geometric and semantic information to infer what container is best to search, or what other objects are best to move, if the target object is hidden from view. We propose to use a 3D scene graph representation to capture the hierarchical, semantic, and geometric aspects of this problem. To exploit this representation in a search process, we introduce Hierarchical Mechanical Search (HMS), a method that guides an agent's actions towards finding a target object specified with a natural language description. HMS is based on a novel neural network architecture that uses neural message passing of vectors with visual, geometric, and linguistic information to allow HMS to reason across layers of the graph while combining semantic and geometric cues. HMS is evaluated on a novel dataset of 500 3D scene graphs with dense placements of semantically related objects in storage locations, and is shown to be significantly better than several baselines at finding objects and close to the oracle policy in terms of the median number of actions required. Additional qualitative results can be found at https: //ai.stanford.edu/mech-search/hms

show abstract

“…Occlusion reasoning in robotics: Object occlusions pose a huge challenge for robot learning: even simple distractors that occasionally occlude the object can cause state-of-the-art RL algorithms to fail [22]. Theoretically, occlusions can be modeled using the framework of POMDP [23], but usually such a formulation is intractable to solve, especially when the states are image observations.…”

Section: Related Workmentioning

confidence: 99%

“…Other than POMDP, there has been lots of work aiming to solve the occlusion problem. Cheng, et al [22] use active vision which learns a policy to move the camera to avoid occlusions. To track the possibly occluded pixels, Ebert, et al [24] use a Conv-LSTM with temporal skip connections to copy pixels from prior images in the history.…”

Section: Related Workmentioning

confidence: 99%

ROLL: Visual Self-Supervised Reinforcement Learning with Object Reasoning

Wang,

Narasimhan,

Lin

et al. 2020

Preprint

View full text Add to dashboard Cite

Current image-based reinforcement learning (RL) algorithms typically operate on the whole image without performing object-level reasoning. This leads to inefficient goal sampling and ineffective reward functions. In this paper, we improve upon previous visual self-supervised RL by incorporating object-level reasoning and occlusion reasoning. Specifically, we use unknown object segmentation to ignore distractors in the scene for better reward computation and goal generation; we further enable occlusion reasoning by employing a novel auxiliary loss and training scheme. We demonstrate that our proposed algorithm, ROLL (Reinforcement learning with Object Level Learning), learns dramatically faster and achieves better final performance compared with previous methods in several simulated visual control tasks. Project video and code are available at https://sites.google.com/andrew.cmu.edu/roll.

show abstract

Reinforcement Learning of Active Vision for Manipulating Objects under Occlusions

Cited by 6 publications

References 22 publications

Active 6D Multi-Object Pose Estimation in Cluttered Scenarios with Deep Reinforcement Learning

Active 6D Multi-Object Pose Estimation in Cluttered Scenarios with Deep Reinforcement Learning

Semantic and Geometric Modeling with Neural Message Passing in 3D Scene Graphs for Hierarchical Mechanical Search

ROLL: Visual Self-Supervised Reinforcement Learning with Object Reasoning

Contact Info

Product

Resources

About