Retrieve segmentation results and populate job parameters

.dockerignore

@@ -1,3 +1,4 @@
 .env
 .git
 .gitignore
+*-env/

.env.example

@@ -47,3 +47,8 @@ USER_PASSWORD=<to-be-specified-per-deployment>
 PREFECT_API_URL=http://prefect:4200/api
 FLOW_NAME="Parent flow/launch_parent_flow"
 TIMEZONE="US/Pacific"
+
+# Environment variables for conda-based Prefect flows
+CONDA_ENV_NAME="dlsia"
+TRAIN_SCRIPT_PATH="src/train.py"
+SEGMENT_SCRIPT_PATH="src/segment.py"

assets/models.json

@@ -1,7 +1,7 @@
 {
     "contents": [
         {
-            "model_name": "DSLIA MSDNet",
+            "model_name": "MSDNet",
             "version": "0.0.1",
             "type": "supervised",
             "user": "mlexchange team",
@@ -48,6 +48,7 @@
                     "name": "dilation_array",
                     "title": "Dilation Array",
                     "param_key": "dilation_array",
+                    "value": "[1, 2, 4]",
                     "placeholder": "e.g. [1, 2, 4]",
                     "error": "Provide a list of ints for dilation",
                     "debounce": 1000,
@@ -230,6 +231,7 @@
                     "name": "weights",
                     "title": "Class Weights",
                     "param_key": "weights",
+                    "value": "[1]",
                     "placeholder": "e.g [0.1, 0.4, 0.5]",
                     "error": "Provide a list with a float for each class",
                     "debounce": 1000,
@@ -312,20 +314,21 @@
                     "title": "Validation %",
                     "param_key": "val_pct",
                     "min": 0,
-                    "max": 100,
-                    "step": 5,
-                    "value": 20,
+                    "max": 1,
+                    "step": 0.05,
+                    "value": 0.2,
+                    "precision": 2,
                     "marks": [
                         {
                             "value": 0,
                             "label": "0%"
                         },
                         {
-                            "value": 50,
+                            "value": 0.5,
                             "label": "50%"
                         },
                         {
-                            "value": 100,
+                            "value": 1,
                             "label": "100%"
                         }
                     ],
@@ -434,7 +437,7 @@
             "reference": "https://dlsia.readthedocs.io/en/latest/"
         },
         {
-            "model_name": "DSLIA TUNet",
+            "model_name": "TUNet",
             "version": "0.0.1",
             "type": "supervised",
             "user": "mlexchange team",
@@ -653,6 +656,7 @@
                     "name": "weights",
                     "title": "Class Weights",
                     "param_key": "weights",
+                    "value": "[1]",
                     "placeholder": "e.g [0.1, 0.4, 0.5]",
                     "error": "Provide a list with a float for each class",
                     "debounce": 1000,
@@ -735,16 +739,21 @@
                     "title": "Validation %",
                     "param_key": "val_pct",
                     "min": 0,
-                    "max": 100,
-                    "step": 5,
-                    "value": 20,
+                    "max": 1,
+                    "step": 0.05,
+                    "value": 0.2,
+                    "precision": 2,
                     "marks": [
                         {
                             "value": 0,
                             "label": "0%"
                         },
                         {
-                            "value": 100,
+                            "value": 0.5,
+                            "label": "50%"
+                        },
+                        {
+                            "value": 1,
                             "label": "100%"
                         }
                     ],
@@ -853,7 +862,7 @@
             "reference": "https://dlsia.readthedocs.io/en/latest/"
         },
         {
-            "model_name": "DSLIA TUNet3+",
+            "model_name": "TUNet3+",
             "version": "0.0.1",
             "type": "supervised",
             "user": "mlexchange team",
@@ -1080,6 +1089,7 @@
                     "name": "weights",
                     "title": "Class Weights",
                     "param_key": "weights",
+                    "value": "[1]",
                     "placeholder": "e.g [0.1, 0.4, 0.5]",
                     "error": "Provide a list with a float for each class",
                     "debounce": 1000,
@@ -1162,16 +1172,21 @@
                     "title": "Validation %",
                     "param_key": "val_pct",
                     "min": 0,
-                    "max": 100,
-                    "step": 5,
-                    "value": 20,
+                    "max": 1,
+                    "step": 0.05,
+                    "value": 0.2,
+                    "precision": 2,
                     "marks": [
                         {
                             "value": 0,
                             "label": "0%"
                         },
                         {
-                            "value": 100,
+                            "value": 0.5,
+                            "label": "50%"
+                        },
+                        {
+                            "value": 1,
                             "label": "100%"
                         }
                     ],

callbacks/control_bar.py

@@ -28,7 +28,7 @@
 from components.parameter_items import ParameterItems
 from constants import ANNOT_ICONS, ANNOT_NOTIFICATION_MSGS, KEY_MODES, KEYBINDS
 from utils.annotations import Annotations
-from utils.data_utils import models, tiled_datasets, tiled_masks, tiled_results
+from utils.data_utils import models, tiled_datasets, tiled_masks
 from utils.plot_utils import generate_notification, generate_notification_bg_icon_col
 
 # TODO - temporary local file path and user for annotation saving and exporting
@@ -230,16 +230,18 @@ def annotation_mode(
             patched_figure["layout"]["dragmode"] = "drawrect"
             annotation_store["dragmode"] = "drawrect"
             styles[trigger] = active
-
         elif trigger == "pan-and-zoom" and pan_and_zoom > 0:
             patched_figure["layout"]["dragmode"] = "pan"
             annotation_store["dragmode"] = "pan"
             styles[trigger] = active
 
     # disable shape editing when in pan/zoom mode
-    for shape in fig["layout"]["shapes"]:
-        shape["editable"] = trigger != "pan-and-zoom" and pan_and_zoom > 0
-    patched_figure["layout"]["shapes"] = fig["layout"]["shapes"]
+    # if no shapes have been added yet,
+    # none need to be set to not editable
+    if "shapes" in fig["layout"]:
+        for shape in fig["layout"]["shapes"]:
+            shape["editable"] = trigger != "pan-and-zoom" and pan_and_zoom > 0
+        patched_figure["layout"]["shapes"] = fig["layout"]["shapes"]
     return (
         patched_figure,
         styles["closed-freeform"],
@@ -853,64 +855,46 @@ def open_controls_drawer(n_clicks, is_opened):
     return no_update, no_update
 
 
+@callback(Output("project-name-src", "data"), Input("refresh-tiled", "n_clicks"))
+def refresh_data_client(refresh_tiled):
+    if refresh_tiled:
+        tiled_datasets.refresh_data_client()
+    data_options = [
+        item for item in tiled_datasets.get_data_project_names() if "seg" not in item
+    ]
+    return data_options
+
+
 @callback(
-    Output("result-selector", "data"),
-    Output("result-selector", "value"),
-    Output("result-selector", "disabled"),
     Output("show-result-overlay-toggle", "checked"),
     Output("show-result-overlay-toggle", "disabled"),
     Output("seg-result-opacity-slider", "disabled"),
-    Output("project-name-src", "data"),
-    Input("project-name-src", "value"),
-    Input("refresh-tiled", "n_clicks"),
     Input("show-result-overlay-toggle", "checked"),
-    State("result-selector", "disabled"),
+    Input("seg-results-train-store", "data"),
+    Input("seg-results-inference-store", "data"),
     State("seg-result-opacity-slider", "disabled"),
 )
-def populate_classification_results(
-    image_src, refresh_tiled, toggle, dropdown_enabled, slider_enabled
+def update_result_controls(
+    toggle, seg_result_train, seg_result_inference, slider_disabled
 ):
-    if refresh_tiled:
-        tiled_datasets.refresh_data_client()
-
-    data_options = [
-        item for item in tiled_datasets.get_data_project_names() if "seg" not in item
-    ]
-    results = []
-    value = None
     checked = False
-    disabled_dropdown = True
-    disabled_toggle = True
-    disabled_slider = True
+    disable_toggle = True
+    disable_slider = True
+    # Disable opacity slider if result overlay is unchecked
     if ctx.triggered_id == "show-result-overlay-toggle":
-        results = no_update
-        value = no_update
         checked = no_update
-        disabled_dropdown = dropdown_enabled
-        disabled_toggle = False
-        disabled_slider = slider_enabled
+        # Must have been enabled to be source of trigger
+        disable_toggle = no_update
+        disable_slider = not slider_disabled
     else:
-        # TODO: Match by mask uid instead of image_src
-        results = [
-            item
-            for item in tiled_results.get_data_project_names()
-            if ("seg" in item and image_src in item)
-        ]
-        if results:
-            value = results[0]
-            disabled_dropdown = False
+        if seg_result_train or seg_result_inference:
             checked = False
-            disabled_toggle = False
-            disabled_slider = False
-
+            disable_toggle = False
+            disable_slider = False
     return (
-        results,
-        value,
-        disabled_dropdown,
         checked,
-        disabled_toggle,
-        disabled_slider,
-        data_options,
+        disable_toggle,
+        disable_slider,
     )
 
 
@@ -961,6 +945,10 @@ def update_model_parameters(model_name):
     ),
 )
 def validate_class_weights(all_annotation_classes, weights):
+
+    if weights is None:
+        return "Provide a list with a float for each class"
+
     parsed_weights = weights.strip("[]").split(",")
     try:
         parsed_weights = [float(weight.strip()) for weight in parsed_weights]
@@ -996,11 +984,17 @@ def validate_class_weights(all_annotation_classes, weights):
     ),
 )
 def validate_dilation_array(dilation_array):
+
+    if dilation_array is None:
+        return "Provide a list of ints for dilation"
+
     parsed_dilation_array = dilation_array.strip("[]").split(",")
     try:
         parsed_dilation_array = [
             int(array_entry.strip()) for array_entry in parsed_dilation_array
         ]
+        if len(parsed_dilation_array) == 0:
+            return "Provide a list of ints for dilation"
         # Check if all elements in the list are floats
         return False
     except ValueError:

callbacks/image_viewer.py

@@ -16,7 +16,7 @@
 from dash.exceptions import PreventUpdate
 
 from constants import ANNOT_ICONS, ANNOT_NOTIFICATION_MSGS, KEYBINDS
-from utils.data_utils import tiled_datasets, tiled_masks, tiled_results
+from utils.data_utils import tiled_datasets, tiled_results
 from utils.plot_utils import (
     create_viewfinder,
     downscale_view,
@@ -70,7 +70,8 @@ def hide_show_segmentation_overlay(toggle_seg_result, opacity):
     State("image-metadata", "data"),
     State("screen-size", "data"),
     State("current-class-selection", "data"),
-    State("result-selector", "value"),
+    State("seg-results-train-store", "data"),
+    State("seg-results-inference-store", "data"),
     State("seg-result-opacity-slider", "value"),
     State("image-viewer", "figure"),
     prevent_initial_call=True,
@@ -85,7 +86,8 @@ def render_image(
     image_metadata,
     screen_size,
     current_color,
-    seg_result_selection,
+    seg_result_train,
+    seg_result_inference,
     opacity,
     fig,
 ):
@@ -118,13 +120,25 @@ def render_image(
                 and ctx.triggered_id == "show-result-overlay-toggle"
             ):
                 return [dash.no_update] * 7 + ["hidden"]
-            annotation_indices = tiled_masks.get_annotated_segmented_results()
-            if str(image_idx + 1) in annotation_indices:
-                # Will not return an error since we already checked if image_idx+1 is in the list
-                mapped_index = annotation_indices.index(str(image_idx + 1))
-                result = tiled_results.get_data_sequence_by_name(seg_result_selection)[
-                    mapped_index
-                ]
+            # Check if the stored results are for the current project and image
+            if seg_result_train or seg_result_inference:
+                seg_result = (
+                    seg_result_inference if seg_result_inference else seg_result_train
+                )
+                if "mask_idx" in seg_result and seg_result["mask_idx"] is not None:
+                    annotation_indices = seg_result["mask_idx"]
+                    if str(image_idx) in annotation_indices:
+                        # Will not return an error since we already checked if image_idx is in the list
+                        mapped_index = annotation_indices.index(str(image_idx))
+                        result = tiled_results.get_data_by_trimmed_uri(
+                            seg_result["seg_result_trimmed_uri"], slice=mapped_index
+                        )
+                    else:
+                        result = None
+                else:
+                    result = tiled_results.get_data_by_trimmed_uri(
+                        seg_result["seg_result_trimmed_uri"], slice=image_idx
+                    )
             else:
                 result = None
     else: