Query-Adapted Metric Learning and Discriminative Feature Selection for Few-Shot Recognition
碩士 === 國立清華大學 === 資訊工程學系所 === 107 === Few-shot recognition aims to recognize novel classes with only a few training samples and alleviates the cost of data collection and labeling. In this thesis, we propose a deep metric and integrate existing metric learning approaches to compare the dissimilarity...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | en_US |
Published: |
2018
|
Online Access: | http://ndltd.ncl.edu.tw/handle/pext4u |
id |
ndltd-TW-107NTHU5392002 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-107NTHU53920022019-05-16T00:52:41Z http://ndltd.ncl.edu.tw/handle/pext4u Query-Adapted Metric Learning and Discriminative Feature Selection for Few-Shot Recognition 應用於小樣本影像辨識之輸入適應之度量學習與具區別性特徵選取 Liou, You-Sin 劉祐欣 碩士 國立清華大學 資訊工程學系所 107 Few-shot recognition aims to recognize novel classes with only a few training samples and alleviates the cost of data collection and labeling. In this thesis, we propose a deep metric and integrate existing metric learning approaches to compare the dissimilarity between each data pair. In addition, we propose a feature selection method by learning a threshold to select discriminative features. Considering that objects vary in scales, we propose a multi-scale feature extractor and include the extracted features in the learned metric to ensure the multi-scale property. Our experiments show that the integration of existing metric learning approaches improves performance on Omniglot dataset and the miniImageNet dataset. Furthermore, experimental results show that our model achieves competitive results to state-of-the-art methods, especially when the amount of training data is more than 1, while using much fewer parameters. Hsu, Chiou-Ting 許秋婷 2018 學位論文 ; thesis 38 en_US |
collection |
NDLTD |
language |
en_US |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 國立清華大學 === 資訊工程學系所 === 107 === Few-shot recognition aims to recognize novel classes with only a few training samples and alleviates the cost of data collection and labeling. In this thesis, we propose a deep metric and integrate existing metric learning approaches to compare the dissimilarity between each data pair. In addition, we propose a feature selection method by learning a threshold to select discriminative features. Considering that objects vary in scales, we propose a multi-scale feature extractor and include the extracted features in the learned metric to ensure the multi-scale property. Our experiments show that the integration of existing metric learning approaches improves performance on Omniglot dataset and the miniImageNet dataset. Furthermore, experimental results show that our model achieves competitive results to state-of-the-art methods, especially when the amount of training data is more than 1, while using much fewer parameters.
|
author2 |
Hsu, Chiou-Ting |
author_facet |
Hsu, Chiou-Ting Liou, You-Sin 劉祐欣 |
author |
Liou, You-Sin 劉祐欣 |
spellingShingle |
Liou, You-Sin 劉祐欣 Query-Adapted Metric Learning and Discriminative Feature Selection for Few-Shot Recognition |
author_sort |
Liou, You-Sin |
title |
Query-Adapted Metric Learning and Discriminative Feature Selection for Few-Shot Recognition |
title_short |
Query-Adapted Metric Learning and Discriminative Feature Selection for Few-Shot Recognition |
title_full |
Query-Adapted Metric Learning and Discriminative Feature Selection for Few-Shot Recognition |
title_fullStr |
Query-Adapted Metric Learning and Discriminative Feature Selection for Few-Shot Recognition |
title_full_unstemmed |
Query-Adapted Metric Learning and Discriminative Feature Selection for Few-Shot Recognition |
title_sort |
query-adapted metric learning and discriminative feature selection for few-shot recognition |
publishDate |
2018 |
url |
http://ndltd.ncl.edu.tw/handle/pext4u |
work_keys_str_mv |
AT liouyousin queryadaptedmetriclearninganddiscriminativefeatureselectionforfewshotrecognition AT liúyòuxīn queryadaptedmetriclearninganddiscriminativefeatureselectionforfewshotrecognition AT liouyousin yīngyòngyúxiǎoyàngběnyǐngxiàngbiànshízhīshūrùshìyīngzhīdùliàngxuéxíyǔjùqūbiéxìngtèzhēngxuǎnqǔ AT liúyòuxīn yīngyòngyúxiǎoyàngběnyǐngxiàngbiànshízhīshūrùshìyīngzhīdùliàngxuéxíyǔjùqūbiéxìngtèzhēngxuǎnqǔ |
_version_ |
1719172640720027648 |