a dirty bbox data in val.json?

[jianqiaomo]

Hi! I found a strange bbox data in: val.json ['annotations'][914]['bbox'] = 50.73, 89.65, 498.14, 0.0

Does the bbox mean "x_left, y_up, w, h"? But why there is a "0.0"? Thanks!

[phamquiluan]

Same problem. I just decided to ignore it when loading annotations :smile:

[zhxgj]

Thank you all for reporting this issue. Let me fix it asap.

[zhxgj]

I have removed the image 342203 from the dev annotation. Please download the revised dev annotation json below. dev.json.zip

[jianqiaomo]

Thank you very much! In fact, I also found some other data:

• in val.json:

[annotations] [ 915 ]['bbox'] = [50.73, 89.65, 498.14, 0.0] [annotations] [ 2511 ]['bbox'] = [308.61, 100.31, 240.1, 0.0] [annotations] [ 18686 ]['bbox'] = [94.53, 284.82, 406.21, 0.0] [annotations] [ 19318 ]['bbox'] = [50.73, 89.65, 498.14, 0.0] [annotations] [ 20288 ]['bbox'] = [51.04, 151.81, 238.11, 0.0] [annotations] [ 20442 ]['bbox'] = [51.04, 320.02, 493.24, 0.0] [annotations] [ 21049 ]['bbox'] = [51.04, 292.81, 493.24, 0.0] [annotations] [ 21171 ]['bbox'] = [51.04, 112.11, 493.24, 0.0] [annotations] [ 34791 ]['bbox'] = [306.14, 215.29, 246.9, 0.0] [annotations] [ 35141 ]['bbox'] = [42.52, 74.9, 246.35, 0.0] [annotations] [ 36214 ]['bbox'] = [308.66, 98.04, 240.2, 0.0] [annotations] [ 43200 ]['bbox'] = [50.73, 416.2, 498.14, 0.0] [annotations] [ 47565 ]['bbox'] = [164.41, 111.79, 510.57, 0.0] [annotations] [ 47595 ]['bbox'] = [101.73, 111.18, 391.81, 0.0] [annotations] [ 64952 ]['bbox'] = [308.66, 89.65, 240.2, 0.0] [annotations] [ 65220 ]['bbox'] = [50.73, 423.2, 498.13, 0.0] [annotations] [ 77743 ]['bbox'] = [286.3, 121.93, 70.87, 0.0]

• in train.json:

[annotations] [ 26038 ] ['bbox'] = [134.76, 68.22, 0.0, 0.0] [annotations] [ 90133 ] ['bbox'] = [47.03, 651.98, 515.54, 0.0] [annotations] [ 150521 ] ['bbox'] = [50.35, 64.0, 510.5, 0.0] [annotations] [ 150531 ] ['bbox'] = [50.35, 64.0, 510.5, 0.0] [annotations] [ 167204 ] ['bbox'] = [62.0, 562.79, 496.0, 0.0] [annotations] [ 243124 ] ['bbox'] = [74.43, 595.22, 472.95, 0.0] [annotations] [ 272739 ] ['bbox'] = [46.49, 675.45, 504.0, 0.0] [annotations] [ 402514 ] ['bbox'] = [60.0, 501.7, 498.0, 0.0] [annotations] [ 423968 ] ['bbox'] = [52.16, 605.14, 490.89, 0.0] [annotations] [ 782584 ] ['bbox'] = [49.04, 624.14, 478.21, 0.0] [annotations] [ 1127619 ] ['bbox'] = [54.42, 93.83, 471.4, 0.0] [annotations] [ 1202730 ] ['bbox'] = [56.69, 495.57, 479.9, 0.0] [annotations] [ 1211904 ] ['bbox'] = [48.0, 652.92, 498.0, 0.0] [annotations] [ 1307067 ] ['bbox'] = [43.6, 566.58, 508.85, 0.0] [annotations] [ 1344996 ] ['bbox'] = [58.88, 567.97, 478.94, 0.0] [annotations] [ 1393908 ] ['bbox'] = [40.72, 666.86, 513.84, 0.0] [annotations] [ 1580676 ] ['bbox'] = [46.49, 573.8, 504.03, 0.0] [annotations] [ 1580688 ] ['bbox'] = [46.52, 639.81, 504.0, 0.0] [annotations] [ 1589957 ] ['bbox'] = [36.85, 663.77, 521.57, 0.0] [annotations] [ 1598952 ] ['bbox'] = [51.02, 596.88, 495.35, 0.0] [annotations] [ 1695215 ] ['bbox'] = [46.0, 681.92, 508.0, 0.0] [annotations] [ 1695382 ] ['bbox'] = [58.0, 600.79, 508.0, 0.0] [annotations] [ 1715796 ] ['bbox'] = [40.5, 623.0, 504.0, 0.0] [annotations] [ 1854303 ] ['bbox'] = [34.11, 577.96, 507.96, 0.0] [annotations] [ 1861141 ] ['bbox'] = [39.0, 659.75, 516.0, 0.0] [annotations] [ 1870418 ] ['bbox'] = [55.14, 119.28, 0.0, 94.14] [annotations] [ 1908434 ] ['bbox'] = [51.68, 512.36, 240.0, 0.0] [annotations] [ 1908449 ] ['bbox'] = [51.68, 597.36, 240.0, 0.0] [annotations] [ 1917572 ] ['bbox'] = [43.94, 498.96, 249.19, 0.0] [annotations] [ 2026931 ] ['bbox'] = [59.84, 656.54, 478.21, 0.0] [annotations] [ 2047182 ] ['bbox'] = [109.29, 643.41, 42.79, 0.0] [annotations] [ 2047186 ] ['bbox'] = [102.83, 640.93, 44.42, 0.0] [annotations] [ 2047210 ] ['bbox'] = [110.92, 649.15, 43.36, 0.0] [annotations] [ 2047214 ] ['bbox'] = [113.75, 649.24, 42.48, 0.0] [annotations] [ 2072330 ] ['bbox'] = [46.49, 530.77, 504.0, 0.0] [annotations] [ 2072567 ] ['bbox'] = [65.2, 619.89, 464.88, 0.0] [annotations] [ 2149187 ] ['bbox'] = [46.49, 655.1, 504.0, 0.0] [annotations] [ 2189686 ] ['bbox'] = [69.73, 517.11, 471.69, 0.0] [annotations] [ 2189768 ] ['bbox'] = [69.73, 581.68, 471.69, 0.0] [annotations] [ 2347748 ] ['bbox'] = [48.31, 88.75, 490.17, 0.0] [annotations] [ 2347767 ] ['bbox'] = [48.31, 577.75, 490.17, 0.0] [annotations] [ 2376011 ] ['bbox'] = [44.79, 665.66, 504.0, 0.0] [annotations] [ 2422653 ] ['bbox'] = [58.59, 541.51, 478.94, 0.0] [annotations] [ 2474524 ] ['bbox'] = [30.74, 630.11, 521.44, 0.0] [annotations] [ 2474528 ] ['bbox'] = [30.74, 647.27, 521.44, 0.0] [annotations] [ 2613360 ] ['bbox'] = [69.73, 545.46, 471.69, 0.0] [annotations] [ 2613531 ] ['bbox'] = [69.73, 597.61, 471.69, 0.0] [annotations] [ 2613617 ] ['bbox'] = [69.73, 628.79, 471.69, 0.0] [annotations] [ 2665702 ] ['bbox'] = [40.72, 609.86, 513.84, 0.0] [annotations] [ 2866969 ] ['bbox'] = [53.86, 651.87, 480.19, 0.0] [annotations] [ 2866979 ] ['bbox'] = [53.86, 652.44, 480.19, 0.0] [annotations] [ 2866991 ] ['bbox'] = [53.86, 651.59, 480.19, 0.0] [annotations] [ 2883891 ] ['bbox'] = [60.95, 369.02, 480.47, 0.0] [annotations] [ 2938869 ] ['bbox'] = [566.78, 94.05, 0.0, 166.24] [annotations] [ 2947003 ] ['bbox'] = [45.0, 630.51, 504.0, 0.0] [annotations] [ 2947011 ] ['bbox'] = [45.0, 610.51, 504.0, 0.0] [annotations] [ 3042870 ] ['bbox'] = [57.09, 508.95, 488.69, 0.0] [annotations] [ 3148454 ] ['bbox'] = [51.68, 592.36, 240.0, 0.0] [annotations] [ 3165283 ] ['bbox'] = [54.0, 604.58, 504.0, 0.0] [annotations] [ 3165441 ] ['bbox'] = [54.0, 637.26, 498.0, 0.0]

However, we can filter them out when we load the annotations. So I think they won't have a great impact on us.

[zhxgj]

Thanks for identifying these problematic bboxes. When you filter them out, I would recommend to filter out the entire page from the dataset, rather than just the bboxes. Because if you only ignore the bbox but keep the page, the page may contain wrong/missed annotations.