Task-Agnostic Object Recognition for Mobile Robots through Few-Shot Image Matching
To assist humans with their daily tasks, mobile robots are expected to navigate complex and dynamic environments, presenting unpredictable combinations of known and unknown objects. Most state-of-the-art object recognition methods are unsuitable for this scenario because they require that: (i) all t...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2020-02-01
|
| Series: | Electronics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2079-9292/9/3/380 |
| _version_ | 1798005981934780416 |
|---|---|
| author | Agnese Chiatti Gianluca Bardaro Emanuele Bastianelli Ilaria Tiddi Prasenjit Mitra Enrico Motta |
| author_facet | Agnese Chiatti Gianluca Bardaro Emanuele Bastianelli Ilaria Tiddi Prasenjit Mitra Enrico Motta |
| author_sort | Agnese Chiatti |
| collection | DOAJ |
| description | To assist humans with their daily tasks, mobile robots are expected to navigate complex and dynamic environments, presenting unpredictable combinations of known and unknown objects. Most state-of-the-art object recognition methods are unsuitable for this scenario because they require that: (i) all target object classes are known beforehand, and (ii) a vast number of training examples is provided for each class. This evidence calls for novel methods to handle unknown object classes, for which fewer images are initially available (few-shot recognition). One way of tackling the problem is learning how to match novel objects to their most similar supporting example. Here, we compare different (shallow and deep) approaches to few-shot image matching on a novel data set, consisting of 2D views of common object types drawn from a combination of ShapeNet and Google. First, we assess if the similarity of objects learned from a combination of ShapeNet and Google can scale up to new object classes, i.e., categories unseen at training time. Furthermore, we show how normalising the learned embeddings can impact the generalisation abilities of the tested methods, in the context of two novel configurations: (i) where the weights of a Convolutional two-branch Network are imprinted and (ii) where the embeddings of a Convolutional Siamese Network are L2-normalised. |
| first_indexed | 2024-04-11T12:48:50Z |
| format | Article |
| id | doaj.art-a37b7617cc3f4a34a79653e2d8b34daa |
| institution | Directory Open Access Journal |
| issn | 2079-9292 |
| language | English |
| last_indexed | 2024-04-11T12:48:50Z |
| publishDate | 2020-02-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Electronics |
| spelling | doaj.art-a37b7617cc3f4a34a79653e2d8b34daa2022-12-22T04:23:17ZengMDPI AGElectronics2079-92922020-02-019338010.3390/electronics9030380electronics9030380Task-Agnostic Object Recognition for Mobile Robots through Few-Shot Image MatchingAgnese Chiatti0Gianluca Bardaro1Emanuele Bastianelli2Ilaria Tiddi3Prasenjit Mitra4Enrico Motta5Knowledge Media Institute, The Open University, Milton Keynes MK7 6AA, UKKnowledge Media Institute, The Open University, Milton Keynes MK7 6AA, UKThe Interaction Lab, Heriot-Watt University, Edinburgh EH14 4AS, UKFaculty of Computer Science, Vrije Universitet Amsterdam, 1081 HV Amsterdam, The NetherlandsInformation Sciences and Technology, The Pennsylvania State University, University Park, PA 16801, USAKnowledge Media Institute, The Open University, Milton Keynes MK7 6AA, UKTo assist humans with their daily tasks, mobile robots are expected to navigate complex and dynamic environments, presenting unpredictable combinations of known and unknown objects. Most state-of-the-art object recognition methods are unsuitable for this scenario because they require that: (i) all target object classes are known beforehand, and (ii) a vast number of training examples is provided for each class. This evidence calls for novel methods to handle unknown object classes, for which fewer images are initially available (few-shot recognition). One way of tackling the problem is learning how to match novel objects to their most similar supporting example. Here, we compare different (shallow and deep) approaches to few-shot image matching on a novel data set, consisting of 2D views of common object types drawn from a combination of ShapeNet and Google. First, we assess if the similarity of objects learned from a combination of ShapeNet and Google can scale up to new object classes, i.e., categories unseen at training time. Furthermore, we show how normalising the learned embeddings can impact the generalisation abilities of the tested methods, in the context of two novel configurations: (i) where the weights of a Convolutional two-branch Network are imprinted and (ii) where the embeddings of a Convolutional Siamese Network are L2-normalised.https://www.mdpi.com/2079-9292/9/3/380few-shot object recognitionimage matchingrobotics |
| spellingShingle | Agnese Chiatti Gianluca Bardaro Emanuele Bastianelli Ilaria Tiddi Prasenjit Mitra Enrico Motta Task-Agnostic Object Recognition for Mobile Robots through Few-Shot Image Matching Electronics few-shot object recognition image matching robotics |
| title | Task-Agnostic Object Recognition for Mobile Robots through Few-Shot Image Matching |
| title_full | Task-Agnostic Object Recognition for Mobile Robots through Few-Shot Image Matching |
| title_fullStr | Task-Agnostic Object Recognition for Mobile Robots through Few-Shot Image Matching |
| title_full_unstemmed | Task-Agnostic Object Recognition for Mobile Robots through Few-Shot Image Matching |
| title_short | Task-Agnostic Object Recognition for Mobile Robots through Few-Shot Image Matching |
| title_sort | task agnostic object recognition for mobile robots through few shot image matching |
| topic | few-shot object recognition image matching robotics |
| url | https://www.mdpi.com/2079-9292/9/3/380 |
| work_keys_str_mv | AT agnesechiatti taskagnosticobjectrecognitionformobilerobotsthroughfewshotimagematching AT gianlucabardaro taskagnosticobjectrecognitionformobilerobotsthroughfewshotimagematching AT emanuelebastianelli taskagnosticobjectrecognitionformobilerobotsthroughfewshotimagematching AT ilariatiddi taskagnosticobjectrecognitionformobilerobotsthroughfewshotimagematching AT prasenjitmitra taskagnosticobjectrecognitionformobilerobotsthroughfewshotimagematching AT enricomotta taskagnosticobjectrecognitionformobilerobotsthroughfewshotimagematching |