Some question about labels

Although I don't know much about cell graphics， it seems that labels No.677-686 don't match with their images very well.

Hi,

We manually check these labels again and do not find issues. Some unlabeled regions you may concern are not cells. Most of them are impurities.

Thanks for your explanation！ But the label No.686 seems compeletely different with the image，which still confuses me.

Hi, Train-Labeled data, cell_00240_label.tiff, Cell id 237, only have one pixel, Is it correct? One-pixel cell label also appear in 244, 267, 268, 296, 501, 751, 768, and 885.

Hi woof,

Thanks for pointing out this issue very much. We apologize for this mistake. The label of case 00686 has been updated just now. https://drive.google.com/file/d/1xRScKh2uF8z6MZTbY1_mZJR0Go-XfyxL/view?usp=sharing

Hi lzxsmy,

Thanks for pointing this out. You can remove these cells. However, when computing the metrics, we do not consider the cells with area < 10 pixels.

We also checked the hidden tuning set labels, there is no cell with area < 10 pixels.