Only center point in 'Annotation' of JSON file

[Jeonous]

Hi.

I had some unexpected segmentation results on the exported JSON file.

it's [128.0, 128.0] in json_file - annotation - [a annotation] - segmentation. it looks like a point, not a polygon.

When I used this JSON file, I got an error with this segmentation value because a polygon with 1 coordinate can't be created.

I thought I made a mistake while working, so I checked the annotations of these images again, but I couldn't find a point in [128.0, 128.0] in images.

I'm using an image with 256 X 256 resolution, so [128.0, 128.0] looks like the center point of the image.

These points appeared on 4-5 images (I worked on a total of 210 images). In some cases, two dots appeared in one image.

I am currently making two different datasets and annotating them, exporting JSON, making a new dataset again, importing these two datasets, merging them, and exporting them again.

Currently, I found the [128.0, 128.0] values and deleted them one by one. I want to find the cause and address this problem.

thank you.

[Jeonous]

Hi,

I used the coco-annotator on 576 X 576 images.

In this case, I found some [288.0, 288.0] points again that I didn't expect (or I didn't annotate) in some annotations (1~2 points per 100 images). Location in JSON:

• 'annotations' - '(integer index value)' - 'segmentation'

It is also the center point of the image.

I still don't know why they appear.

After finding them, I deleted them in the extracted JSON file, not on the application. I cant' find the point in my annotation records.

They make errors during deep learning model training since it's impossible to generate a polygon with only 1 point.