MLT (Multi-Lingual) 2017 论文 | 下载链接
注意:在下载之前,请先注册一个账号。
MLT 2017 数据集包含两个任务:任务 1 是文本检测 (多语言文本)。 任务2是文本识别。
有11个与任务1相关的文件需要下载,它们分别是:
ch8_training_images_x.zip(x from 1 to 8)
ch8_validation_images.zip
ch8_training_localization_transcription_gt_v2.zip
ch8_validation_localization_transcription_gt_v2.zip
测试集不需要下载。
有6个与任务2相关的文件需要下载,它们分别是:
ch8_training_word_images_gt_part_x.zip (x from 1 to 3)
ch8_validation_word_images_gt.zip
ch8_training_word_gt_v2.zip
ch8_validation_word_gt_v2.zip
在下载完成后, 将文件放于 [path-to-data-dir]
文件夹内,如下所示:
path-to-data-dir/
mlt2017/
# text detection
ch8_training_images_1.zip
ch8_training_images_2.zip
ch8_training_images_3.zip
ch8_training_images_4.zip
ch8_training_images_5.zip
ch8_training_images_6.zip
ch8_training_images_7.zip
ch8_training_images_8.zip
ch8_training_localization_transcription_gt_v2.zip
ch8_validation_images.zip
ch8_validation_localization_transcription_gt_v2.zip
# word recognition
ch8_training_word_images_gt_part_1.zip
ch8_training_word_images_gt_part_2.zip
ch8_training_word_images_gt_part_3.zip
ch8_training_word_gt_v2.zip
ch8_validation_word_images_gt.zip
ch8_validation_word_gt_v2.zip