-
Notifications
You must be signed in to change notification settings - Fork 13
Description
Hi,
I’m trying to get DeepD3 running and ran into an issue with the GUI.
What I’m doing
Dataset: x23 (channel 1) from the Model Zoo.
I can load the stack and scroll through Z in the GUI.
Then I go to Analyze → Select model → Find… and choose DeepD3_32F.h5.
Problem
After I click Open, the model path does not appear in the text box (it still says “Find…”), and if I click Save and start segmenting, the dialog just closes and nothing happens (no output, no overlay).
What I tried
If I install TensorFlow with pip install tensorflow (latest), DeepD3 won’t start at all.
If I pin versions (e.g., TensorFlow 2.13.x / Keras 2.13.x), the GUI starts, the stack loads, but the model still isn’t registered by the dialog.
Environment
OS: Windows [10/11] (64-bit)
Python: [3.10.x] (Anaconda/conda env)
DeepD3: [pip show deepd3 → version]
TensorFlow: [2.13.1 or other]
Keras: [2.13.x or other]
NumPy: [1.24.x or other]
PyTables (tables): [3.9.x or other]
Do you have guidance on which TensorFlow/Keras/NumPy versions are currently recommended for the GUI model selector, or any known workaround (e.g., accepted Keras .h5 formats, specific versions, or a way to enter the model path manually)?