debug print

src/htr_pipeline/inferencer.py

@@ -26,25 +26,51 @@ class Inferencer:
 
     @timer_func
     def predict_regions(self, input_image, pred_score_threshold=0.5, containments_threshold=0.5, visualize=True):
+        import time
+
+        t1 = time.time()
+
         input_image = self.preprocess_img.binarize_img(input_image)
 
         image = mmcv.imread(input_image)
+
+        t2 = time.time()
+
+        print(f"Function executed bin and read in {(t2-t1):.4f}s")
+
+        t1 = time.time()
+
         result = self.seg_model(image, return_datasample=True)
         result_pred = result["predictions"][0]
+        t2 = time.time()
+
+        print(f"Function executed predict in {(t2-t1):.4f}s")
+
+        t1 = time.time()
 
         filtered_result_pred = self.postprocess_seg_mask.filter_on_pred_threshold(
             result_pred, pred_score_threshold=pred_score_threshold
         )
 
+        t2 = time.time()
+
+        print(f"Function executed filter in {(t2-t1):.4f}s")
+
         if len(filtered_result_pred.pred_instances.masks) == 0:
             raise gr.Error("No Regions were predicted by the model")
 
         else:
+            t1 = time.time()
+
             result_align = self.process_seg_mask.align_masks_with_image(filtered_result_pred, image)
             result_clean = self.postprocess_seg_mask.remove_overlapping_masks(
                 predicted_mask=result_align, containments_threshold=containments_threshold
             )
 
+            t2 = time.time()
+
+            print(f"Function executed align and remove in {(t2-t1):.4f}s")
+
             if visualize:
                 result_viz = self.seg_model.visualize(
                     inputs=[image], preds=[result_clean], return_vis=True, no_save_vis=True
@@ -52,6 +78,8 @@ class Inferencer:
             else:
                 result_viz = None
 
+            t1 = time.time()
+
             regions_cropped, polygons = self.process_seg_mask.crop_masks(result_clean, image)
             order = self.ordering.order_regions_marginalia(result_clean)
 
@@ -59,6 +87,10 @@ class Inferencer:
             polygons_ordered = [polygons[i] for i in order]
             masks_ordered = [result_clean.pred_instances.masks[i] for i in order]
 
+            t2 = time.time()
+
+            print(f"Function executed crop and margin in {(t2-t1):.4f}s")
+
             return result_viz, regions_cropped_ordered, polygons_ordered, masks_ordered
 
     @timer_func

src/htr_pipeline/utils/pipeline_inferencer.py

@@ -6,6 +6,8 @@ from src.htr_pipeline.utils.xml_helper import XMLHelper
 
 terminate = False
 
+# TODO check why region is so slow to start.. Is their error with loading the model?
+
 
 class PipelineInferencer:
     def __init__(self, process_seg_mask: SegMaskHelper, xml_helper: XMLHelper):
@@ -29,7 +31,7 @@ class PipelineInferencer:
         pred_score_threshold_regions,
         pred_score_threshold_lines,
         containments_threshold,
-        htr_threshold=0.7,
+        htr_threshold=0.6,
     ):
         global terminate
 
@@ -77,7 +79,7 @@ class PipelineInferencer:
         return region_data if mean_htr_score > htr_threshold + 0.1 else None
 
     def _process_lines(
-        self, text_region, inferencer, pred_score_threshold, containments_threshold, mask, region_id, htr_threshold=0.7
+        self, text_region, inferencer, pred_score_threshold, containments_threshold, mask, region_id, htr_threshold=0.6
     ):
         _, lines_cropped_ordered, line_polygons_ordered = inferencer.predict_lines(
             text_region, pred_score_threshold, containments_threshold, visualize=False, custom_track=False

tabs/htr_tool.py

@@ -86,7 +86,7 @@ with gr.Blocks() as htr_tool_tab:
                                     )
                                 with gr.Row():
                                     gr.Slider(
-                                        value=0.7,
+                                        value=0.6,
                                         minimum=0.5,
                                         maximum=1,
                                         label="HTR threshold",
@@ -94,7 +94,7 @@ with gr.Blocks() as htr_tool_tab:
                                         scale=1,
                                     )
                                     gr.Slider(
-                                        value=0.8,
+                                        value=0.7,
                                         minimum=0.6,
                                         maximum=1,
                                         label="Avg threshold",
@@ -105,7 +105,7 @@ with gr.Blocks() as htr_tool_tab:
                                 htr_tool_region_segment_model_dropdown = gr.Dropdown(
                                     choices=["Riksarkivet/RmtDet_region"],
                                     value="Riksarkivet/RmtDet_region",
-                                    label="Region segment model",
+                                    label="Region Segment models",
                                     info="Will add more models later!",
                                 )
 
@@ -113,15 +113,15 @@ with gr.Blocks() as htr_tool_tab:
                                 htr_tool_line_segment_model_dropdown = gr.Dropdown(
                                     choices=["Riksarkivet/RmtDet_lines"],
                                     value="Riksarkivet/RmtDet_lines",
-                                    label="Line segment model",
+                                    label="Line Segment models",
                                     info="Will add more models later!",
                                 )
 
                                 htr_tool_transcriber_model_dropdown = gr.Dropdown(
                                     choices=["Riksarkivet/SATRN_transcriber", "microsoft/trocr-base-handwritten"],
                                     value="Riksarkivet/SATRN_transcriber",
-                                    label="Transcribe model",
-                                    info="Will add more models later!",
+                                    label="Transcriber models",
+                                    info="Models will be continuously  updated with future additions for specific cases.",
                                 )
 
                     with gr.Column(scale=2):