How to evaluate the image gradient in new models after using human-in-the-loop

[Crudeoil0]

Hi I am using cellpose to fit into my cardiac cell and I am wondering how to evaluate the image gradient in newly trained models after using human-in-the-loop.

[carsen-stringer]

do you mean the flows? and by evaluate what do you mean? you can use your model on the command line with python -m cellpose --dir /path/to/images --pretrained_model name_in_gui

[Crudeoil0]

Thank you for your comment, Carsen. I am a biologist at the University of Auckland, a new hand in machine learning. Thank you for your useful tool. It helps a lot in my research. The cell, according to your paper, seems to generate seeds in the center and calculate the gradient around it to detect the boundary. What I wish to know is how to get the gradient on the human-in-the-loop trained model.

[carsen-stringer]

you can get it in the same way, the flows are saved in the _seg.npy file as a dictionary key, after you run the new model