Bug In Dataset Generator

When I generate a dataset, the number of images reported to have been generated is very different to the actual number of images in the downloaded dataset (See images below). Furthermore, when I go to train an Ultralytics Yolov8 segmentation model on the generated dataset, it fails immediately. with the following error:

Epoch GPU_mem box_loss seg_loss cls_loss dfl_loss Instances Size
0%| | 0/119 [00:00<?, ?it/s]…/aten/src/ATen/native/cuda/IndexKernel.cu:92: operator(): block: [391,0,0], thread: [64,0,0] Assertion -sizes[i] <= index && index < sizes[i] && "index out of bounds" failed.

I have attempted to troubleshoot this issue by checking the following:

  1. All images have a corresponding label file
  2. Seg labels are within correct range 0<val<1
  3. Images can be opened/read by openCV without error
  4. Class indices are within the expected range 0<labelIndex<labelCount

Any help with this is appreciated.

Project Type: Segmentation
Operating System & Browser: Windows, Edge

Reported Number of Generated Images

Number of Generated Images When Viewed

Actual Number of Images When Downloaded

Can you try generating the dataset again and see if the error persists.

Thanks for the reply. That seems to have done the trick

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.