Add nice widgets callbacks

plantseg/resources/models_zoo.yaml

@@ -8,7 +8,7 @@ generic_confocal_3D_unet:
   description: "Unet trained on confocal images of Arabidopsis Ovules on 1/2-resolution in XY with BCEDiceLoss."
   dimensionality: "3D"
   modality: "confocal"
-  recommended_patch_size: [80, 170, 170]
+  recommended_patch_size: [ 80, 160, 160 ]
   output_type: "boundaries"
 
 
@@ -18,7 +18,7 @@ generic_light_sheet_3D_unet:
   description: "Unet trained on light-sheet images of Arabidopsis Lateral Root Primordia on original resolution with BCEDiceLoss."
   dimensionality: "3D"
   modality: "light-sheet"
-  recommended_patch_size: [ 80, 170, 170 ]
+  recommended_patch_size: [ 80, 160, 160 ]
   output_type: "boundaries"
 
 ## Ovules
@@ -28,7 +28,7 @@ confocal_3D_unet_ovules_ds1x:
   description: "Unet trained on confocal images of Arabidopsis Ovules on original resolution with BCEDiceLoss."
   dimensionality: "3D"
   modality: "confocal"
-  recommended_patch_size: [ 80, 170, 170 ]
+  recommended_patch_size: [ 80, 160, 160 ]
   output_type: "boundaries"
 
 confocal_3D_unet_ovules_ds2x:
@@ -37,7 +37,7 @@ confocal_3D_unet_ovules_ds2x:
   description: "Unet trained on confocal images of Arabidopsis Ovules on 1/2-resolution in XY with BCEDiceLoss."
   dimensionality: "3D"
   modality: "confocal"
-  recommended_patch_size: [ 80, 170, 170 ]
+  recommended_patch_size: [ 80, 160, 160 ]
   output_type: "boundaries"
 
 confocal_3D_unet_ovules_ds3x:
@@ -46,7 +46,7 @@ confocal_3D_unet_ovules_ds3x:
   description: "Unet trained on confocal images of Arabidopsis Ovules on 1/3-resolution in XY with BCEDiceLoss."
   dimensionality: "3D"
   modality: "confocal"
-  recommended_patch_size: [ 80, 170, 170 ]
+  recommended_patch_size: [ 80, 160, 160 ]
   output_type: "boundaries"
 
 confocal_2D_unet_ovules_ds2x:
@@ -55,7 +55,7 @@ confocal_2D_unet_ovules_ds2x:
   description: "2D Unet trained on z-slices of confocal images of Arabidopsis Ovules (1/2-resolution in XY) with BCEDiceLoss."
   dimensionality: "2D"
   modality: "confocal"
-  recommended_patch_size: [ 80, 170, 170 ]
+  recommended_patch_size: [ 1, 256, 256 ]
   output_type: "boundaries"
 
 confocal_3D_unet_ovules_nuclei_ds1x:
@@ -64,7 +64,7 @@ confocal_3D_unet_ovules_nuclei_ds1x:
   description: "Unet trained on confocal images of Arabidopsis Ovules nuclei stain on original resolution with BCEDiceLoss. The network predicts 1 channel: nuclei probability maps"
   dimensionality: "3D"
   modality: "confocal"
-  recommended_patch_size: [ 80, 170, 170 ]
+  recommended_patch_size: [ 80, 160, 160 ]
   output_type: "nuclei"
 
 ## Root
@@ -74,7 +74,7 @@ lightsheet_3D_unet_root_ds1x:
   description: "Unet trained on light-sheet images of Lateral Root Primordia on original resolution with BCEDiceLoss."
   dimensionality: "3D"
   modality: "light-sheet"
-  recommended_patch_size: [ 80, 170, 170 ]
+  recommended_patch_size: [ 80, 160, 160 ]
   output_type: "boundaries"
 
 lightsheet_3D_unet_root_ds2x:
@@ -83,7 +83,7 @@ lightsheet_3D_unet_root_ds2x:
   description: "Unet trained on light-sheet images of Lateral Root Primordia on 1/2-resolution in XY with BCEDiceLoss."
   dimensionality: "3D"
   modality: "light-sheet"
-  recommended_patch_size: [ 80, 170, 170 ]
+  recommended_patch_size: [ 80, 160, 160 ]
   output_type: "boundaries"
 
 lightsheet_3D_unet_root_ds3x:
@@ -92,7 +92,7 @@ lightsheet_3D_unet_root_ds3x:
   description: "Unet trained on light-sheet images of Lateral Root Primordia on 1/3-resolution in XY with BCEDiceLoss."
   dimensionality: "3D"
   modality: "light-sheet"
-  recommended_patch_size: [ 80, 170, 170 ]
+  recommended_patch_size: [ 80, 160, 160 ]
   output_type: "boundaries"
 
 lightsheet_2D_unet_root_ds1x:
@@ -101,7 +101,7 @@ lightsheet_2D_unet_root_ds1x:
   description: "2D Unet trained on z-slices of light-sheet images of Lateral Root Primordia on original resolution with BCEDiceLoss."
   dimensionality: "2D"
   modality: "light-sheet"
-  recommended_patch_size: [ 80, 170, 170 ]
+  recommended_patch_size: [ 1, 256, 256 ]
   output_type: "boundaries"
 
 lightsheet_3D_unet_root_nuclei_ds1x:
@@ -110,7 +110,7 @@ lightsheet_3D_unet_root_nuclei_ds1x:
   description: "Unet trained on light-sheet images of Lateral Root Primordia nuclei on original resolution with BCEDiceLoss. The network predicts 2 channels: nuclei mask in the 1st channel, nuclei boundaries in the 2nd channel"
   dimensionality: "3D"
   modality: "light-sheet"
-  recommended_patch_size: [ 80, 170, 170 ]
+  recommended_patch_size: [ 80, 160, 160 ]
   output_type: "nuclei"
 
 lightsheet_2D_unet_root_nuclei_ds1x:
@@ -119,7 +119,7 @@ lightsheet_2D_unet_root_nuclei_ds1x:
   description: "2D Unet trained on z-slices of light-sheet images of Lateral Root Primordia nuclei on original resolution with BCEDiceLoss. The network predicts 2 channels: nuclei mask in the 1st channel, nuclei boundaries in the 2nd channel"
   dimensionality: "2D"
   modality: "light-sheet"
-  recommended_patch_size: [ 80, 170, 170 ]
+  recommended_patch_size: [ 1, 256, 256 ]
   output_type: "nuclei"
 
 # PNAS
@@ -138,7 +138,7 @@ confocal_3D_unet_sa_meristem_cells:
   description: "3D Unet trained on confocal images of Arabidopsis thaliana apical stem cell: https://www.repository.cam.ac.uk/handle/1810/262530"
   dimensionality: "3D"
   modality: "confocal"
-  recommended_patch_size: [ 80, 170, 170 ]
+  recommended_patch_size: [ 80, 160, 160 ]
   output_type: "boundaries"
 
 # Mouse embryo ex vivo
@@ -148,7 +148,7 @@ lightsheet_3D_unet_mouse_embryo_cells:
   description: "A a variant of 3D U-Net trained to predict the cell boundaries in live light-sheet images of ex-vivo developing mouse embryo. Voxel size: (0.2×0.2×1 µm^3) (XYZ)"
   dimensionality: "3D"
   modality: "light-sheet"
-  recommended_patch_size: [ 80, 170, 170 ]
+  recommended_patch_size: [ 80, 160, 160 ]
   output_type: "boundaries"
 
 confocal_3D_unet_mouse_embryo_nuclei:

plantseg/run_plantseg.py

@@ -15,6 +15,8 @@ def parser():
 
 
 def main():
+    from plantseg.utils import check_version
+    check_version()
     args = parser()
 
     if args.gui:

plantseg/utils.py

@@ -4,11 +4,13 @@
 from pathlib import Path
 from shutil import copy2
 from typing import Tuple, Optional
+from warnings import warn
 
 import requests
 import yaml
 
 from plantseg import model_zoo_path, custom_zoo, home_path, PLANTSEG_MODELS_DIR, plantseg_global_path
+from plantseg.__version__ import __version__ as current_version
 from plantseg.pipeline import gui_logger
 
 CONFIG_TRAIN_YAML = "config_train.yml"
@@ -276,3 +278,36 @@ def clean_models():
 
         else:
             print("Invalid input, please type 'y' or 'n'.")
+
+
+def check_version(plantseg_url=' https://api.github.com/repos/hci-unihd/plant-seg/releases/latest'):
+    try:
+        response = requests.get(plantseg_url).json()
+        latest_version = response['tag_name']
+
+    except requests.exceptions.ConnectionError:
+        warn("Connection error, could not check for new version.")
+        return None
+    except requests.exceptions.Timeout:
+        warn("Connection timeout, could not check for new version.")
+        return None
+    except requests.exceptions.TooManyRedirects:
+        warn("Too many redirects, could not check for new version.")
+        return None
+    except Exception as e:
+        warn(f"Unknown error, could not check for new version. Error: {e}")
+        return None
+
+    latest_version_numeric = [int(x) for x in latest_version.split(".")]
+    plantseg_version_numeric = [int(x) for x in current_version.split(".")]
+
+    if len(latest_version_numeric) != len(plantseg_version_numeric):
+        warn(f"Could not check for new version, version number not in the correct format.\n"
+             f"Current version: {current_version}, latest version: {latest_version}")
+        return None
+
+    for l_v, p_v in zip(latest_version_numeric, plantseg_version_numeric):
+        if l_v > p_v:
+            print(f"New version of PlantSeg available: {latest_version}.\n"
+                  f"Please update your version to the latest one!")
+            return None

plantseg/viewer/headless.py

@@ -10,10 +10,12 @@
 from plantseg.viewer.dag_handler import DagHandler
 from plantseg.viewer.widget.predictions import ALL_DEVICES, ALL_CUDA_DEVICES
 
-all_gpus_str = f'all {len(ALL_CUDA_DEVICES)} gpus'
+all_gpus_str = f'all gpus: {len(ALL_CUDA_DEVICES)}'
 ALL_GPUS = [all_gpus_str] if len(ALL_CUDA_DEVICES) > 0 else []
 ALL_DEVICES_HEADLESS = ALL_DEVICES + ALL_GPUS
 
+MAX_WORKERS = len(ALL_CUDA_DEVICES) if len(ALL_CUDA_DEVICES) > 0 else multiprocessing.cpu_count()
+
 
 def _parse_input_paths(inputs, path_suffix='_path'):
     list_input_paths = [_input for _input in inputs if _input[-len(path_suffix):] == path_suffix]
@@ -39,7 +41,7 @@ def run_workflow_headless(path):
                            'widget_type': 'IntSlider',
                            'tooltip': 'Define the size of the gaussian smoothing kernel. '
                                       'The larger the more blurred will be the output image.',
-                           'max': multiprocessing.cpu_count(), 'min': 1},
+                           'max': MAX_WORKERS, 'min': 1},
               scheduler={'label': 'Scheduler',
                          'choices': ['multiprocessing', 'threaded']
                          },
@@ -48,7 +50,7 @@ def run_workflow_headless(path):
     def run(list_inputs: input_hints,
             out_directory: Path = Path.home(),
             device: str = ALL_DEVICES_HEADLESS[0],
-            num_workers: int = 1,
+            num_workers: int = MAX_WORKERS,
             scheduler: str = 'multiprocessing'):
         dict_of_jobs = {}
         cluster = distributed.LocalCluster(n_workers=num_workers, threads_per_worker=1)