YOLOv5在训练数据的时，出现标签corrupted，标签的坐标出现了负值。

问题描述：在使用YOLO进行训练数据时，出现了Scanning 'C:\Users\aoqia\Desktop\datasets\grape\train\labels' images and labels...0 found, 4320 missing, 0 empty, 0 corrupted:这个错误。具体如下所示。

 WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000002_1.jpg: negative labels
train: WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000002_2.jpg: negative labels
train: WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000022_2.jpg: non-normalized or out of bounds coordinate labels
train: WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000022_4.jpg: non-normalized or out of bounds coordinate labels
train: WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000022_5.jpg: non-normalized or out of bounds coordinate labels
train: WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000022_9.jpg: non-normalized or out of bounds coordinate labels
train: WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000025_7.jpg: negative labels
train: WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000025_9.jpg: negative labels
train: WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000031_3.jpg: non-normalized or out of bounds coordinate labels
train: WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000031_4.jpg: non-normalized or out of bounds coordinate labels
train: WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000031_9.jpg: non-normalized or out of bounds coordinate labels
train: WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000039_4.jpg: non-normalized or out of bounds coordinate labels
train: WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000039_9.jpg: non-normalized or out of bounds coordinate labels
train: WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000095_2.jpg: negative labels
train: WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000098_3.jpg: negative labels
train: WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000098_4.jpg: negative labels
train: WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000098_6.jpg: negative labels
train: WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000111_3.jpg: negative labels
train: WARNING: Ignoring corrupted image and/or label C:\Users\aoqia\Desktop\datasets\grape\train\images\000116_3.jpg: negative labels

原因分析：我们阅读报错提示，发现数据集标签中矩形框的坐标出现了负值，查看报错提示的标签，确实出现了负值，比如下图中的倒数第二行中出现了负值。这是因为我们在使用数据扩增的时候，使用了平移扩增，迫使边缘部分的矩形框被截掉，出现矩形框坐标为负值。

 解决办法：在使用数据扩增的时候取消使用平移扩增（或者说是图片缩放截取扩增）。