pretrained hdf5 file for U-Net_2D_ZeroCostDL4Mic

[jazberna1]

Hello,

I just started using the notebook U-Net_2D_ZeroCostDL4Mic.ipynb. I have excuted the section "1.1. Load key dependencies" and then I moved straight to the section "5.1 Generate prediction(s) from unseen dataset". There I indicate Data_folder (where I have my images) and Results_folder. My question is, for the Prediction_model_folder, is there somewhere an already trained hdf5 file that I can use to directly try to segment my images? or otherwise I must always train the model myself first.

Thanks Jorge

[jazberna1]

Hi @IvanHCenalmor ,

Apologies, I went through the notebook and I actually says in order to use a model previously generated model I must run Sections 1, 2 and 6. Should it be 5 instead of 6 (Version log)?

I am able to run section 1. However in section 2 I am unclear to what I must supply in model_name and model_path of section 2. From the notebook I read:

model_name: Use only my_model -style, not my-model. If you want to use a previously trained model, enter the name of the pretrained model (which should be contained in the trained_model -folder after training).

model_path: Enter the path of the folder where you want to save your model.

In my case I have available already a weights_best.hdf5 file which I just want to use further down in the notebook to generate predicition(s) from an unseen dataset.

Any help will be grateful. Thanks Jorge