search

cvat-ai / cvat

Annotate better with CVAT, the industry-leading data engine for machine learning. Used and trusted by teams at any scale, for data of any scale.
https://cvat.ai
MIT License
12.73k stars 3.02k forks source link

YOLOv8 segmentation annotations export error #8739

Closed fattypiggy closed 6 days ago

fattypiggy commented 6 days ago

Actions before raising this issue

Steps to Reproduce

  1. Go to project
  2. click the three dots
  3. export dataset
  4. format as "YOLOv8 segmentation 1.0"
  5. used the exported annotations and images to draw the masks back for double-checking.

Here are three images, the first one is the original image to annotate, the second is the exported result, and the third is the expected result annotated in the CVAT.ai website(I have the membership). As you can see, the exported result only contains two classes, but actually, the ground truth should contain 5 classes. And this is just an example, my dataset contains 60 images, and most of them miss either classes or instances.

FYI: I annotated the images with the mix of polygon and mask, and it worked well last month, actually the same result exported from last month is the same as what I annotated, but they are different right now.

14_2 14_2_mask anontations from CVAT

Expected Behavior

I drew the result exported on Nov 30th, it worked well. 14_1_mask

Possible Solution

No response

Context

14_2.txt 14_2 You can write a visualization program yourself to validate using the 14_2.txt and the image below.

Environment

cvat.ai website.
fattypiggy commented 6 days ago

my fault, fixed already.