Convpaint Usage

Step-by-Step Protocol for Using Convpaint

Convpaint Usage in Napari: 

1. Open your data:

• Launch Napari from the terminal:

conda activate napari_env

napari

• Open your image via File > Open.

2. Open Convpaint:

• Go to Plugins > Convpaint.

• The Convpaint panel will appear in the interface.

3. Create annotations:

• Add an annotation layer to create a labels layer.

• Draw your annotations: 

• Use the brush tool to paint regions. 

• If you want to detect one type of object, use: 

• Label 1: background

• Label 2: foreground

4. Image type and normalization: 

• Select the correct image type: 

• Single channel image

• Multichannel image

• RGB image

• Choose a normalization method: 

• No normalization

• Normalize over stack

• Normalize by plane

5. Train and segment:

• Click Train to train the pixel classifier

• Click Segment Image to apply the model.

• If Auto segment is enabled, segmentation runs automatically after training.

• If segmentation is not accurate: 

• Add more annotations. 

• Click Train again. 

• Re-run segmentation.

6. Convert binary segmentation into instances:

• To convert the binary segmentation into individual object instances, install the plugin inside Napari: 

napari-segment-blobs-and-things-with-membranes

• Open the plugin from: 

• Tools > Segment blobs and things with membranes

• Select: 

Connected Component labeling. 

• Choose your binary segmentation image and run. 

• A new labels will be created where each connected object has an independent label ID.

7. Save the model:

• Click Save model. 

• You have two options: 

• .yml file: saves only the model settings.

• .pkl file: saves the full trained model.

8. Reload a model: 

• Click Load Model.

• Select a previously saved .yml or .pkl file.