mlt2017.md

MLT 2017 Datasets

Data Downloading

MLT (Multi-Lingual) 2017 dataset Paper | Download Link

Note: Please register an account to download this dataset.

MLT 2017 dataset consists of two tasks. Task 1 is Text detection (Multi-Language Script) and Task 2 is Word Recognition.

Text Detection(Multi-script)

The 11 files downloaded for task 1 are

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

No need to download the Test Set.

Word Identification

The 6 files downloaded for task 2 are

 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

After downloading the files, place them under [path-to-data-dir] folder:

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

Back to dataset converters