New bbm

backend/projects/annotation_registry.py

@@ -108,6 +108,15 @@
             "type": "labels",
         },
     },
+    "OCRTextlineSegmentation": {
+        "ocr_transcribed_json": {
+            "to_name": "image_url",
+            "from_name": [
+                "annotation_bboxes",
+            ],
+            "type": ["rectangle"],
+        },
+    },
     "OCRTranscription": {
         "ocr_transcribed_json": {
             "to_name": "image_url",
@@ -190,59 +199,6 @@
 }
 
 
-def convert_prediction_json_to_annotation_result(
-    prediction_json, speakers_json, audio_duration, index, is_acoustic=False
-):
-    """
-    Convert prediction_json and transcribed_json to annotation_result
-    """
-
-    result = []
-    if prediction_json == None:
-        return result
-
-    for idx, val in enumerate(prediction_json):
-        label_dict = {
-            "origin": "manual",
-            "to_name": "audio_url",
-            "from_name": "labels",
-            "original_length": audio_duration,
-        }
-        text_dict = {
-            "origin": "manual",
-            "to_name": "audio_url",
-            "from_name": "transcribed_json",
-            "original_length": audio_duration,
-        }
-        if is_acoustic:
-            text_dict["from_name"] = "verbatim_transcribed_json"
-        id = f"shoonya_{index}s{idx}s{generate_random_string(13-len(str(idx)))}"
-        label_dict["id"] = id
-        text_dict["id"] = id
-        label_dict["type"] = "labels"
-        text_dict["type"] = "textarea"
-
-        value_labels = {
-            "start": val["start"],
-            "end": val["end"],
-            "labels": [
-                next(
-                    speaker
-                    for speaker in speakers_json
-                    if speaker["speaker_id"] == val["speaker_id"]
-                )["name"]
-            ],
-        }
-        value_text = {"start": val["start"], "end": val["end"], "text": [val["text"]]}
-
-        label_dict["value"] = value_labels
-        text_dict["value"] = value_text
-        result.append(label_dict)
-        result.append(text_dict)
-
-    return result
-
-
 def convert_conversation_json_to_annotation_result(conversation_json, idx):
     result = []
     for i in range(len(conversation_json)):
@@ -262,12 +218,15 @@ def convert_conversation_json_to_annotation_result(conversation_json, idx):
 
 
 def draft_data_json_to_annotation_result(draft_data_json, project_type, pk=None):
+    from projects.views import convert_prediction_json_to_annotation_result
+
     registry_helper = ProjectRegistry.get_instance()
     input_dataset_info = registry_helper.get_input_dataset_and_fields(project_type)
     dataset_model = getattr(dataset_models, input_dataset_info["dataset_type"])
     try:
         dataset_item = dataset_model.objects.get(pk=pk)
     except:
+        dataset_item = None
         pass
     result = []
     idx = 0
@@ -286,13 +245,20 @@ def draft_data_json_to_annotation_result(draft_data_json, project_type, pk=None)
             if field == "conversation_json":
                 ans = convert_conversation_json_to_annotation_result(value, idx)
             elif field == "transcribed_json" or field == "prediction_json":
-                ans = convert_prediction_json_to_annotation_result(
+                assert type(value) in [
+                    list,
+                    dict,
+                ], f"Something wrong is there in the type of {value}"
+                if isinstance(value, list):
+                    value = {"verbatim_transcribed_json": value}
+                sub_ans = convert_prediction_json_to_annotation_result(
+                    None,
+                    project_type,
+                    dataset_item,
                     value,
-                    dataset_item.speakers_json,
-                    dataset_item.audio_duration,
-                    idx,
-                    project_type == "AcousticNormalisedTranscriptionEditing",
+                    True,
                 )
+                ans.extend(sub_ans)
             else:
                 if field_type == "textarea":
                     field_dict["value"] = {"text": [value]}

backend/projects/label_studio_jsx_files/ocr/ocr_textline_segmentation.jsx

@@ -0,0 +1,5 @@
+<View>
+  <Style>.ant-input { font-size: large; }</Style>
+  <Image name="image_url" value="$image_url"/>
+  <Rectangle name="annotation_bboxes" toName="image_url" strokeWidth="3" className="ignore_assertion"/>
+</View>

backend/projects/project_registry.yaml

@@ -99,6 +99,24 @@ OCR:
         fields:
           annotations:
             - ocr_transcribed_json
+    OCRTextlineSegmentation:
+      project_mode: "Annotation"
+      label_studio_jsx_file: "ocr/ocr_textline_segmentation.jsx"
+      input_dataset:
+        class: OCRDocument
+        fields:
+          - image_url
+          - page_number
+        display_fields:
+          - image_url
+          - page_number
+        prediction: ocr_prediction_json
+      output_dataset: 
+        class: OCRDocument
+        save_type: in_place
+        fields:
+          annotations:
+            - ocr_transcribed_json
     OCRTranscriptionEditing:
       project_mode: "Annotation"
       label_studio_jsx_file: "ocr/ocr_transcription.jsx"

backend/projects/utils.py

@@ -25,6 +25,7 @@
 import yaml
 from yaml.loader import SafeLoader
 from jiwer import wer
+from dataset import models as dataset_models
 from users.utils import generate_random_string
 from utils.convert_result_to_chitralekha_format import (
     create_memory,
@@ -367,6 +368,7 @@ def process_ocr_tasks(
     task,
     is_OCRSegmentCategorization,
     is_OCRSegmentCategorizationEditing,
+    is_OCRTextlineSegmentation,
     is_OCRSegmentCategorisationRelationMappingEditing,
 ):
     annotation_result = process_annotation_result(task)
@@ -375,6 +377,7 @@ def process_ocr_tasks(
         annotation_result,
         is_OCRSegmentCategorization,
         is_OCRSegmentCategorizationEditing,
+        is_OCRTextlineSegmentation,
         is_OCRSegmentCategorisationRelationMappingEditing,
     )
 
@@ -474,6 +477,7 @@ def process_ocr_results(
     annotation_result,
     is_OCRSegmentCategorization,
     is_OCRSegmentCategorizationEditing,
+    is_OCRTextlineSegmentation,
     is_OCRSegmentCategorisationRelationMappingEditing,
 ):
     from projects.views import convert_annotation_result_to_formatted_json
@@ -482,12 +486,15 @@ def process_ocr_results(
         annotation_result,
         None,
         False,
-        is_OCRSegmentCategorization or is_OCRSegmentCategorizationEditing,
+        is_OCRSegmentCategorization
+        or is_OCRSegmentCategorizationEditing
+        or is_OCRTextlineSegmentation,
         False,
     )
     if (
         is_OCRSegmentCategorization
         or is_OCRSegmentCategorizationEditing
+        or is_OCRTextlineSegmentation
         or is_OCRSegmentCategorisationRelationMappingEditing
     ):
         bboxes_relation_json = []
@@ -513,6 +520,7 @@ def process_task(
     include_input_data_metadata_json,
     dataset_model,
     is_audio_project_type,
+    fetch_parent_data_field,
 ):
     task_dict = model_to_dict(task)
     if export_type != "JSON":
@@ -547,6 +555,21 @@ def process_task(
         task_dict["data"]["input_data_metadata_json"] = dataset_model.objects.get(
             pk=task_dict["input_data"]
         ).metadata_json
+    try:
+        if fetch_parent_data_field and dataset_model:
+            parent_data_item = dataset_model.objects.get(
+                pk=task_dict["input_data"]
+            ).parent_data
+            if parent_data_item:
+                dataset_model = getattr(
+                    dataset_models, parent_data_item.instance_id.dataset_type
+                )
+                parent_dataset_model = dataset_model.objects.get(pk=parent_data_item.id)
+                task_dict["data"]["fetch_parent_data_field"] = getattr(
+                    parent_dataset_model, fetch_parent_data_field, None
+                )
+    except Exception as e:
+        pass
 
     del task_dict["annotation_users"]
     del task_dict["review_user"]