Low-shot Visual Recognition
Many real world datasets are characterized by having a long tailed distribution, with several samples for some classes and only a few samples for other classes. While many Deep Learning based solutions exist for object recognition when hundreds of samples are available, there are not many solutions...
Main Author: | |
---|---|
Other Authors: | |
Format: | Others |
Published: |
Virginia Tech
2016
|
Subjects: | |
Online Access: | http://hdl.handle.net/10919/73321 |
id |
ndltd-VTETD-oai-vtechworks.lib.vt.edu-10919-73321 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-VTETD-oai-vtechworks.lib.vt.edu-10919-733212021-10-21T05:32:54Z Low-shot Visual Recognition Pemula, Latha Electrical and Computer Engineering Batra, Dhruv Parikh, Devi Abbott, A. Lynn Visual Recognition Object Recognition Computer Vision Low-shot Learning Many real world datasets are characterized by having a long tailed distribution, with several samples for some classes and only a few samples for other classes. While many Deep Learning based solutions exist for object recognition when hundreds of samples are available, there are not many solutions for the case when there are only a few samples available per class. Recognition in the regime where the number of training samples available for each class are low, ranging from 1 to couple of tens of examples is called Lowshot Recognition. In this work, we attempt to solve this problem. Our framework is similar to [1]. We use a related dataset with sufficient number (a couple of hundred) of samples per class to learn representations using a Convolutional Neural Network (CNN). This CNN is used to extract features of the lowshot samples and learn a classifier . During representation learning, we enforce the learnt representations to obey certain property by using a custom loss function. We believe that when the lowshot sample obey this property the classification step becomes easier. We show that the proposed solution performs better than the softmax classifier by a good margin. Master of Science 2016-10-25T08:00:37Z 2016-10-25T08:00:37Z 2016-10-24 Thesis vt_gsexam:9048 http://hdl.handle.net/10919/73321 In Copyright http://rightsstatements.org/vocab/InC/1.0/ ETD application/pdf Virginia Tech |
collection |
NDLTD |
format |
Others
|
sources |
NDLTD |
topic |
Visual Recognition Object Recognition Computer Vision Low-shot Learning |
spellingShingle |
Visual Recognition Object Recognition Computer Vision Low-shot Learning Pemula, Latha Low-shot Visual Recognition |
description |
Many real world datasets are characterized by having a long tailed distribution, with several samples for some classes and only a few samples for other classes. While many Deep Learning based solutions exist for object recognition when hundreds of samples are available, there are not many solutions for the case when there are only a few samples available per class. Recognition in the regime where the number of training samples available for each class are low, ranging from 1 to couple of tens of examples is called Lowshot Recognition. In this work, we attempt to solve this problem. Our framework is similar to [1]. We use a related dataset with sufficient number (a couple of hundred) of samples per class to learn representations using a Convolutional Neural Network (CNN). This CNN is used to extract features of the lowshot samples and learn a classifier . During representation learning, we enforce the learnt representations to obey certain property by using a custom loss function. We believe that when the lowshot sample obey this property the classification step becomes easier. We show that the proposed solution performs better than the softmax classifier by a good margin. === Master of Science |
author2 |
Electrical and Computer Engineering |
author_facet |
Electrical and Computer Engineering Pemula, Latha |
author |
Pemula, Latha |
author_sort |
Pemula, Latha |
title |
Low-shot Visual Recognition |
title_short |
Low-shot Visual Recognition |
title_full |
Low-shot Visual Recognition |
title_fullStr |
Low-shot Visual Recognition |
title_full_unstemmed |
Low-shot Visual Recognition |
title_sort |
low-shot visual recognition |
publisher |
Virginia Tech |
publishDate |
2016 |
url |
http://hdl.handle.net/10919/73321 |
work_keys_str_mv |
AT pemulalatha lowshotvisualrecognition |
_version_ |
1719490918346653696 |