Hi everyone 👋

First off, thank you for sharing this amazing project! 🙏
I'm currently exploring this repository Chin-Fu-Liu/Acute-stroke_Detection_Segmentation and would like to use the trained models in a PyTorch environment.

I noticed that the trained weights are saved in .h5 format (which I assume were trained using Keras/TensorFlow). I have two main questions:

1. 
   

   Is there a way to convert the existing .h5 weights to PyTorch format?

   
   
   
   
   • 
     

     If yes, could you please guide me or point to any scripts that might help with the conversion?

     
     
   
   
   
   
   
2. 
   

   Alternatively, can you share the full architecture (e.g., in PyTorch or a model summary)?

   
   
   
   
   • 
     

     This would help in recreating the same architecture in PyTorch so I can manually port over weights (if needed).

     
     
   
   
   
   
   

I’d be happy to contribute a PyTorch implementation back to the repo if that's helpful. 😊

Thanks in advance!