Hello all,
I hope this gets fixed quickly, as it literally was working less than 24 hours ago.
I have two datasets of around 7k images each: lm/t2swir7kend-2 and lm/t2_s2_swir_0_7k .
I have tried to export my dataset after generating it.
When it comes to the zip, I get random “chunks” of that dataset; this is reflected also in your “view all images” section of the version just created.
Example:
You can see “train: 6290”, same as “dataset” size on the left, but effectively “1857 total images”, than another time 3700, another time 2k etc, completely randomic.
I have tried to duplicate, create instance segmentation project instead of object detection, whatever. This resulted in a very big loss of time and a very big loss of credits on my account unfortunately.
I hope a solution can be found as the re-carefully manually labelled images are now stuck on my workspace. Thank you in advance for any kind of help!
I am also having this issue as well as many others it appears. I made a post where you can see others having this issue. It worked fine 1/24 but the next version from 1/25 started to have this issue.
I have the same issue. I guess the deduplication is malfunctioning. I tried multiple times (spent money) and every time I got a different number of images in the downloaded dataset.
My name is Ford and I’m a Support Engineer at Roboflow. Thank you for flagging this issue, and we apologize for the inconvenience this has caused.
We identified an issue causing only a fraction of images to download when exporting dataset versions. We have deployed a fix for this issue. To apply the fix, please generate a new dataset version.
Please let us know in this thread if you’re still experiencing incomplete downloads after generating a new dataset version, and we are happy to help further.
Again, we apologize for the inconvenience this issue has caused, and thank you for your patience and understanding!
