Scanning 'bottle_data/train/labelTxt' images and labels...16229 found, 23 missing, 1 empty, 466 corrupted:

[LULU-LULULA]

（单目标识别）使用自己的数据集进行训练，使用roLabelImg标注目标为XML格式，再转为DOTA的四点坐标生成txt文件，在进行训练时显示 23 missing, 1 empty, 466 corrupted，可是数据集之前可视化过，图片和标注信息均无损坏，请问该怎么解决啊。
train: Scanning 'bottle_data/train/labelTxt' images and labels...16229 found, 23 missing, 1 empty, 466 corrupted: 100%|████████████████████████████████████████████████████████████████████| 16252/16252 [00:02<00:00, 7261.77it/s] train: WARNING: bottle_data/train/images/1_Sand_000004_79_2.png: ignoring corrupt image/label: negative label values [ -0.5 -0.5], please check your dota format labels train: WARNING: bottle_data/train/images/1_Sand_000004_93_2.png: ignoring corrupt image/label: negative label values [ -0.5 -0.5 -0.5 -0.5], please check your dota format labels train: WARNING: bottle_data/train/images/1_Sand_000005_85_1.png: ignoring corrupt image/label: negative label values [ -0.6 -0.5], please check your dota format labels train: WARNING: bottle_data/train/images/1_Sand_000007_38_1.png: ignoring corrupt image/label: negative label values [ -0.5 -0.5], please check your dota format labels train: WARNING: bottle_data/train/images/1_Sand_000007_40_1.png: ignoring corrupt image/label: negative label values [ -0.5 -0.5], please check your dota format labels train: WARNING: bottle_data/train/images/1_Sand_000009_91_2.png: ignoring corrupt image/label: negative label values [ -0.4 -0.5 -0.5], please check your dota format labels train: WARNING: bottle_data/train/images/1_Sand_000011_77_2.png: ignoring corrupt image/label: negative label values [ -0.5 -0.5], please check your dota format labels train: WARNING: bottle_data/train/images/1_Sand_000011_85_1.png: ignoring corrupt image/label: negative label values [ -0.5 -0.5], please check your dota format labels train: WARNING: bottle_data/train/images/1_Sand_000012_55_1.png: ignoring corrupt image/label: negative label values [ -0.1], please check your dota format labels
问题图片：

代码说出错的图片1_Sand_000001_79_2.png

对应txt文件1_Sand_000001_79_2.txt

[LULU-LULULA]

问题已解决。是标签转化中由于被切割存在负值，只要改成0就好

[pangyanhua]

请问您是怎么修改的啊，我也遇到了同样的问题

[LULU-LULULA]

我在知乎上写了解决的帖子，你搜搜这个问题，应该可以搜到。 | | LULU露露 | | ***@***.*** |

…

---- 回复的原邮件 ---- | 发件人 | Pang yan ***@***.***> | | 日期 | 2022年11月14日 11:34 | | 收件人 | ***@***.***> | | 抄送至 | ***@***.******@***.***> | | 主题 | Re: [hukaixuan19970627/yolov5_obb] Scanning 'bottle_data/train/labelTxt' images and labels...16229 found, 23 missing, 1 empty, 466 corrupted: (Issue #478) | 请问您是怎么修改的啊，我也遇到了同样的问题 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>