method: hoainv2020-12-25

Authors: Nguyen Viet Hoai

Affiliation: HUST

Description: Model used model EfficientNet-B5 as feature extractor from image and model Transformer as sequence model. We used LabelSmoothing in loss function