The datalist are used in text spotting experiments:
- cocotext.json
- icdar2013_train_datalist.json
- icdar2013_test_datalist.json
- icdar2015_train_datalist.json
- icdar2015_test_datalist.json
- icdar2017_train_datalist.json
- icdar2017_test_datalist.json
- icdar2019_train_datalist.json
- icdar2019-art_train_datalist.json
- scut-foru_train_datalist.json
- total_text_train_datalist.json
- total_text_test_datalist.json
- ctw1500_train_datalist_with_cares.json
- ctw1500_test_datalist_with_cares.json
- synthtext_80w.json
- synthtext_with_char.json
- train_syntext_word_eng.json
- train_emcs_imgs.json
All formatted datalist can be downloaded from Link (Access Code：5JBf)
All datalists are transformed into a unified Davar Format like:
{
"Images/train/img19.jpg": {
"height": 1280,
"width": 960,
"content_ann": {
"bboxes": [[145, 202, 322, 246, 480, 333, 449, 380, 288, 306, 114, 263],
[519, 338, 695, 468, 788, 591, 740, 618, 615, 496, 473, 394]],
"cares": [1, 1],
"texts": ["Chateau", "Fujisawa"]
}
},
"Images/train/img1102.jpg": {
"height": 1259,
"width": 1280,
"content_ann": {
"bboxes": [[798, 984, 861, 984, 923, 984, 923, 1050, 861, 1050, 798, 1050],
[398, 984, 450, 984, 502, 984, 502, 1060, 450, 1060, 398, 1060],
[334, 466, 546, 313, 792, 373, 845, 457, 764, 522, 692, 453, 560, 431, 422, 528, 420, 531],
[532, 528, 641, 528, 643, 665, 554, 666],
[422, 729, 580, 810, 770, 731, 838, 758, 614, 889, 373, 778]],
"cares": [1, 1, 0, 1, 1],
"texts": ["AID", "FI", "###", "6", "PATTAYA"]
}
},
...
}