Feature hallucination in hypersphere space for few‐shot classification
Abstract Few‐shot classification (FSC) targeting at classifying unseen classes with few labelled samples is still a challenging task. Recent works show that transfer‐learning based approaches are competitive with meta‐learning ones, which usually pre‐train a convolutional neural networks (CNN)‐based...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2022-11-01
|
| Series: | IET Image Processing |
| Online Access: | https://doi.org/10.1049/ipr2.12579 |
| _version_ | 1828340431768780800 |
|---|---|
| author | Sai Yang Fan Liu Zhiyu Chen |
| author_facet | Sai Yang Fan Liu Zhiyu Chen |
| author_sort | Sai Yang |
| collection | DOAJ |
| description | Abstract Few‐shot classification (FSC) targeting at classifying unseen classes with few labelled samples is still a challenging task. Recent works show that transfer‐learning based approaches are competitive with meta‐learning ones, which usually pre‐train a convolutional neural networks (CNN)‐based network using cross‐entropy (CE) loss and throw away the last layer to post‐process the novel classes. Hereby, they still suffer the issue of getting a more transferable extractor and lacking enough labelled novel samples. Thus, the authors propose the algorithm of feature hallucination in hypersphere space (FHHS) for FSC. On the first stage, the authors pre‐train a more transferable feature extractor using a hypersphere loss (HL), which supplies CE with supervised contrastive (SC) loss and self‐supervised loss (SSL), in which SC can map the base and novel images onto the hypersphere space densely. On the second stage, the authors generate new samples for unseen classes using their novel algorithm of synthetic novel sampling with the base (SNSB), which linearly interpolate between each novel class prototype and its K nearest neighbour base class prototypes. Comprehensive experiments on multiple popular FSC demonstrate that HL loss can enhance the performance of backbone network and the authors’ feature hallucination method is superior to the existing hallucination‐based methods. |
| first_indexed | 2024-04-13T22:56:25Z |
| format | Article |
| id | doaj.art-27189c89dc9a4fae996fdae215ee4acc |
| institution | Directory Open Access Journal |
| issn | 1751-9659 1751-9667 |
| language | English |
| last_indexed | 2024-04-13T22:56:25Z |
| publishDate | 2022-11-01 |
| publisher | Wiley |
| record_format | Article |
| series | IET Image Processing |
| spelling | doaj.art-27189c89dc9a4fae996fdae215ee4acc2022-12-22T02:25:59ZengWileyIET Image Processing1751-96591751-96672022-11-0116133603361610.1049/ipr2.12579Feature hallucination in hypersphere space for few‐shot classificationSai Yang0Fan Liu1Zhiyu Chen2School of Electrical Engineering Nantong University Nantong Jiangsu ChinaCollege of Computer and Information Hohai University Nanjing Jiangsu ChinaCollege of Computer and Information Hohai University Nanjing Jiangsu ChinaAbstract Few‐shot classification (FSC) targeting at classifying unseen classes with few labelled samples is still a challenging task. Recent works show that transfer‐learning based approaches are competitive with meta‐learning ones, which usually pre‐train a convolutional neural networks (CNN)‐based network using cross‐entropy (CE) loss and throw away the last layer to post‐process the novel classes. Hereby, they still suffer the issue of getting a more transferable extractor and lacking enough labelled novel samples. Thus, the authors propose the algorithm of feature hallucination in hypersphere space (FHHS) for FSC. On the first stage, the authors pre‐train a more transferable feature extractor using a hypersphere loss (HL), which supplies CE with supervised contrastive (SC) loss and self‐supervised loss (SSL), in which SC can map the base and novel images onto the hypersphere space densely. On the second stage, the authors generate new samples for unseen classes using their novel algorithm of synthetic novel sampling with the base (SNSB), which linearly interpolate between each novel class prototype and its K nearest neighbour base class prototypes. Comprehensive experiments on multiple popular FSC demonstrate that HL loss can enhance the performance of backbone network and the authors’ feature hallucination method is superior to the existing hallucination‐based methods.https://doi.org/10.1049/ipr2.12579 |
| spellingShingle | Sai Yang Fan Liu Zhiyu Chen Feature hallucination in hypersphere space for few‐shot classification IET Image Processing |
| title | Feature hallucination in hypersphere space for few‐shot classification |
| title_full | Feature hallucination in hypersphere space for few‐shot classification |
| title_fullStr | Feature hallucination in hypersphere space for few‐shot classification |
| title_full_unstemmed | Feature hallucination in hypersphere space for few‐shot classification |
| title_short | Feature hallucination in hypersphere space for few‐shot classification |
| title_sort | feature hallucination in hypersphere space for few shot classification |
| url | https://doi.org/10.1049/ipr2.12579 |
| work_keys_str_mv | AT saiyang featurehallucinationinhyperspherespaceforfewshotclassification AT fanliu featurehallucinationinhyperspherespaceforfewshotclassification AT zhiyuchen featurehallucinationinhyperspherespaceforfewshotclassification |