2018 Fiscal Year Research-status Report
A Sequence-to-sequence Model based Dissimilarity Measurement for Clustering Structural Data
| Project/Area Number |
18K18068
|
|---|
| Research Institution | Tokyo University of Agriculture and Technology |
|---|
Principal Investigator |
NGUYENTUAN CUONG 東京農工大学, 工学(系)研究科(研究院), 特任助教 (10814246)
|
|---|
| Project Period (FY) |
2018-04-01 – 2021-03-31
|
| Keywords | clustering / weakly supervised / hierarchical features / CNN |
|---|
| Outline of Annual Research Achievements |
We proposed a CNN based method to learn both localization and classification representations of mathematical symbols in handwritten formula images. Symbols in various scales are located and classified by multi-level features of multi-scaled CNN. We train the CNN networks by weakly supervised training and fine-tune them by symbols attention to enhance classification and location prediction. Multi-level spatial representations are extracted from the CNN for calculating the distance. Experiments on our collected datasets and the CROHME dataset show the promising results.
|
| Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
We have proposed a learning model for clustering offline handwritten mathematical expression.
|
| Strategy for Future Research Activity |
We continue to develop the method for clustering online handwritten mathematical expression. We focus on weakly supervised learning method and hierarchical representation.
|
| Causes of Carryover |
We are going to use the budget for attending international conferences, revising and publishing journal papers.
|
Research Products
(10 results)
