TY - GEN
T1 - Heterogeneous data integration using confidence estimation of unseen visual data for zero-shot learning
AU - Seo, Sanghyun
AU - Kim, Juntae
N1 - Publisher Copyright:
© 2019 Association for Computing Machinery.
PY - 2019/1/10
Y1 - 2019/1/10
N2 - Zero-shot learning is a learning methodology that can be used to recognize concepts that have never been seen during the training phase. Recently, interest in zero-shot learning has been increased by embedding multi-modal data into common vector space through heterogeneous data integration methodology. However, since the existing methodologies compare heterogeneous data focusing on the similarity between each vector, the performance of zero-shot learning decreases when the number of semantic candidates increases. We propose a heterogeneous data integration methodology using a confidence estimator for unseen visual data which estimates that whether input data is unseen data or not and output confidence measure. The proposed methodology constructs a more efficient zero-shot learning model by applying estimated confidence of input unseen visual data to the visual-semantic distance obtained from heterogeneous data integration model. Experiments have shown that the proposed methodology can improve zero-shot learning performance for unseen data despite a small performance decrease in the seen data.
AB - Zero-shot learning is a learning methodology that can be used to recognize concepts that have never been seen during the training phase. Recently, interest in zero-shot learning has been increased by embedding multi-modal data into common vector space through heterogeneous data integration methodology. However, since the existing methodologies compare heterogeneous data focusing on the similarity between each vector, the performance of zero-shot learning decreases when the number of semantic candidates increases. We propose a heterogeneous data integration methodology using a confidence estimator for unseen visual data which estimates that whether input data is unseen data or not and output confidence measure. The proposed methodology constructs a more efficient zero-shot learning model by applying estimated confidence of input unseen visual data to the visual-semantic distance obtained from heterogeneous data integration model. Experiments have shown that the proposed methodology can improve zero-shot learning performance for unseen data despite a small performance decrease in the seen data.
KW - Confidence estimation
KW - Cross modal retrieval
KW - Heterogeneous data integration
KW - Visual semantic embedding
KW - Zero-shot learning
UR - http://www.scopus.com/inward/record.url?scp=85063576643&partnerID=8YFLogxK
U2 - 10.1145/3305160.3305216
DO - 10.1145/3305160.3305216
M3 - Conference contribution
AN - SCOPUS:85063576643
T3 - ACM International Conference Proceeding Series
SP - 171
EP - 174
BT - Proceedings of the 2019 2nd International Conference on Software Engineering and Information Management, ICSIM 2019 - Workshop 2019 2nd International Conference on Big Data and Smart Computing, ICBDSC 2019
PB - Association for Computing Machinery
T2 - 2nd International Conference on Software Engineering and Information Management, ICSIM 2019 - and its Workshop 2019 2nd International Conference on Big Data and Smart Computing, ICBDSC 2019
Y2 - 10 January 2019 through 13 January 2019
ER -