Libtorch_UNet_inQT

• QT Project
• UNet
• Libtorch (Pytorch C API)
• Deep Leaning, Image Segmentation

Version

• OS : Windows10
• VS : VS2019
• OpenCV : 4.5.2
• CUDA : 10.2
• Libtorch : CUDA 10.2 (Release, Debug)

Usage Example

I used the MVTEC anomaly detection dataset.
Link : https://www.mvtec.com/company/research/datasets/mvtec-ad/

Dataset folder organization

• dataset
  • carpet
    • train
      • image
      • mask
    • val
      • image
      • mask

You have to match the name of the mask and image file as follows.

• in image folder
  • somename_0.png (or .jpg, .bmp, ...)
  • somename_1.png
  • somename_2.png
  • ...
• in mask folder
  • somename_0_mask.png (or .jpg, .bmp, ...)
  • somename_1_mask.png
  • somename_2_mask.png
  • ...

Dataset Load

Labeling Tool

Train

Test