Title | ||
---|---|---|
Deep transformer: A framework for 2D text image rectification from planar transformations. |
Abstract | ||
---|---|---|
•This paper proposed the first planar text image rectification method using deep neural network which uses 2-stage supervised spatial neural network.•The model requires milder assumptions on a few parallel text lines in text area.•A new dataset with Chinese charset is collected to test this algorithm.•Different model configurations are tested, compared and analyzed on the collected dataset.•Inner mechanism is analyzed and derived to show why supervised canonical information is helpful in rectification. |
Year | DOI | Venue |
---|---|---|
2018 | 10.1016/j.neucom.2018.02.015 | Neurocomputing |
Keywords | Field | DocType |
DNN,Image rectification,Image understanding | Architecture,Pattern recognition,Image rectification,Segmentation,Transformer,Robustness (computer science),Geometric transformation,Planar,Artificial intelligence,Artificial neural network,Mathematics | Journal |
Volume | ISSN | Citations |
289 | 0925-2312 | 0 |
PageRank | References | Authors |
0.34 | 31 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Chengzhe Yan | 1 | 0 | 0.68 |
Jie Hu | 2 | 20 | 1.45 |
Changshui Zhang | 3 | 5506 | 323.40 |