fix: LEAP-1692: Image export for COCO and YOLO (#6855)

Co-authored-by: Sergei Ivashchenko <[email protected]> Co-authored-by: triklozoid <[email protected]> Co-authored-by: MihajloHoma <[email protected]>
HumanSignal · Jan 21, 2025 · b38000e · b38000e
1 parent 40f156a
commit b38000e
Show file tree

Hide file tree

Showing 7 changed files with 529 additions and 70 deletions.
diff --git a/label_studio/data_export/api.py b/label_studio/data_export/api.py
@@ -204,7 +204,7 @@ def get(self, request, *args, **kwargs):
         logger.debug('Prepare export files')
 
         export_file, content_type, filename = DataExport.generate_export_file(
-            project, tasks, export_type, download_resources, request.GET
+            project, tasks, export_type, download_resources, request.GET, hostname=request.build_absolute_uri('/')
         )
 
         r = FileResponse(export_file, as_attachment=True, content_type=content_type, filename=filename)
@@ -569,7 +569,7 @@ def get(self, request, *args, **kwargs):
             return response
 
 
-def async_convert(converted_format_id, export_type, project, **kwargs):
+def async_convert(converted_format_id, export_type, project, hostname, download_resources=False, **kwargs):
     with transaction.atomic():
         try:
             converted_format = ConvertedFormat.objects.get(id=converted_format_id)
@@ -583,7 +583,7 @@ def async_convert(converted_format_id, export_type, project, **kwargs):
         converted_format.save(update_fields=['status'])
 
     snapshot = converted_format.export
-    converted_file = snapshot.convert_file(export_type)
+    converted_file = snapshot.convert_file(export_type, download_resources=download_resources, hostname=hostname)
     if converted_file is None:
         raise ValidationError('No converted file found, probably there are no annotations in the export snapshot')
     md5 = Export.eval_md5(converted_file)
@@ -645,6 +645,7 @@ def post(self, request, *args, **kwargs):
         serializer = ExportConvertSerializer(data=request.data, context={'project': snapshot.project})
         serializer.is_valid(raise_exception=True)
         export_type = serializer.validated_data['export_type']
+        download_resources = serializer.validated_data.get('download_resources')
 
         with transaction.atomic():
             converted_format, created = ConvertedFormat.objects.get_or_create(export=snapshot, export_type=export_type)
@@ -657,6 +658,8 @@ def post(self, request, *args, **kwargs):
             converted_format.id,
             export_type,
             snapshot.project,
+            request.build_absolute_uri('/'),
+            download_resources=download_resources,
             on_failure=set_convert_background_failure,
         )
         return Response({'export_type': export_type, 'converted_format': converted_format.id})
diff --git a/label_studio/data_export/mixins.py b/label_studio/data_export/mixins.py
@@ -137,6 +137,9 @@ def _get_export_serializer_option(serialization_options):
                 options['context']['interpolate_key_frames'] = serialization_options['interpolate_key_frames']
             if serialization_options.get('include_annotation_history') is False:
                 options['omit'] = ['annotations.history']
+            # download resources
+            if serialization_options.get('download_resources') is True:
+                options['download_resources'] = True
         return options
 
     def get_task_queryset(self, ids, annotation_filter_options):
@@ -303,7 +306,7 @@ def run_file_exporting(self, task_filter_options=None, annotation_filter_options
                 serialization_options=serialization_options,
             )
 
-    def convert_file(self, to_format):
+    def convert_file(self, to_format, download_resources=False, hostname=None):
         with get_temp_dir() as tmp_dir:
             OUT = 'out'
             out_dir = pathlib.Path(tmp_dir) / OUT
@@ -313,7 +316,10 @@ def convert_file(self, to_format):
                 config=self.project.get_parsed_config(),
                 project_dir=None,
                 upload_dir=out_dir,
-                download_resources=False,
+                download_resources=download_resources,
+                # for downloading resource we need access to the API
+                access_token=self.project.organization.created_by.auth_token.key,
+                hostname=hostname,
             )
             input_name = pathlib.Path(self.file.name).name
             input_file_path = pathlib.Path(tmp_dir) / input_name

diff --git a/label_studio/data_export/models.py b/label_studio/data_export/models.py
@@ -142,7 +142,7 @@ def get_export_formats(project):
         return sorted(formats, key=lambda f: f.get('disabled', False))
 
     @staticmethod
-    def generate_export_file(project, tasks, output_format, download_resources, get_args):
+    def generate_export_file(project, tasks, output_format, download_resources, get_args, hostname=None):
         """Generate export file and return it as an open file object.
 
         Be sure to close the file after using it, to avoid wasting disk space.
@@ -161,6 +161,8 @@ def generate_export_file(project, tasks, output_format, download_resources, get_
             project_dir=None,
             upload_dir=os.path.join(settings.MEDIA_ROOT, settings.UPLOAD_DIR),
             download_resources=download_resources,
+            access_token=project.organization.created_by.auth_token.key,
+            hostname=hostname,
         )
         with get_temp_dir() as tmp_dir:
             converter.convert(input_json, tmp_dir, output_format, is_dir=False)

diff --git a/label_studio/data_export/serializers.py b/label_studio/data_export/serializers.py
@@ -166,6 +166,7 @@ class SerializationOption(serializers.Serializer):
 
 class ExportConvertSerializer(serializers.Serializer):
     export_type = serializers.CharField(help_text='Export file format.')
+    download_resources = serializers.BooleanField(help_text='Download resources in converter.', required=False)
 
     def validate_export_type(self, value):
         project = self.context.get('project')

diff --git a/label_studio/tests/export.tavern.yml b/label_studio/tests/export.tavern.yml
@@ -112,67 +112,116 @@ stages:
     url: '{django_live_url}/api/projects/{pk}/export/formats'
   response:
     json:
-      - title: 'JSON'
-        description: !anystr
-        link: 'https://labelstud.io/guide/export.html#JSON'
-        name: 'JSON'
-      - title: 'JSON-MIN'
-        description: !anystr
-        link: 'https://labelstud.io/guide/export.html#JSON-MIN'
-        name: 'JSON_MIN'
-      - title: 'CSV'
-        description: !anystr
-        link: 'https://labelstud.io/guide/export.html#CSV'
-        name: 'CSV'
-      - title: 'TSV'
-        description: !anystr
-        link: 'https://labelstud.io/guide/export.html#TSV'
-        name: 'TSV'
-      - title: 'COCO'
-        description: 'Popular machine learning format used by the COCO dataset for object detection and image segmentation tasks with polygons and rectangles.'
-        link: 'https://labelstud.io/guide/export.html#COCO'
-        tags: ['image segmentation', 'object detection']
-        name: 'COCO'
-      - title: 'YOLO'
-        description: 'Popular TXT format is created for each image file. Each txt file contains annotations for the corresponding image file, that is object class, object coordinates, height & width.'
-        link: 'https://labelstud.io/guide/export.html#YOLO'
-        tags: ['image segmentation', 'object detection']
-        name: 'YOLO'
-      - title: 'YOLOv8 OBB'
-        description: 'Popular TXT format is created for each image file. Each txt file contains annotations for the corresponding image file. The YOLO OBB format designates bounding boxes by their four corner points with coordinates normalized between 0 and 1, so it is possible to export rotated objects.'
-        link: 'https://labelstud.io/guide/export.html#YOLO'
-        tags: ['image segmentation', 'object detection']
-        name: 'YOLO_OBB'
-      - title: 'CONLL2003'
-        description: 'Popular format used for the CoNLL-2003 named entity recognition challenge.'
-        link: 'https://labelstud.io/guide/export.html#CONLL2003'
-        tags: ['sequence labeling', 'text tagging', 'named entity recognition']
-        name: 'CONLL2003'
+      - title: JSON
+        description: List of items in raw JSON format stored in one JSON file. Use to export
+          both the data and the annotations for a dataset. It's Label Studio Common Format
+        link: https://labelstud.io/guide/export.html#JSON
+        name: JSON
+      - title: JSON-MIN
+        description: List of items where only "from_name", "to_name" values from the raw
+          JSON format are exported. Use to export only the annotations for a dataset.
+        link: https://labelstud.io/guide/export.html#JSON-MIN
+        name: JSON_MIN
+      - title: CSV
+        description: Results are stored as comma-separated values with the column names
+          specified by the values of the "from_name" and "to_name" fields.
+        link: https://labelstud.io/guide/export.html#CSV
+        name: CSV
+      - title: TSV
+        description: Results are stored in tab-separated tabular file with column names
+          specified by "from_name" "to_name" values
+        link: https://labelstud.io/guide/export.html#TSV
+        name: TSV
+      - title: COCO
+        description: Popular machine learning format used by the COCO dataset for object
+          detection and image segmentation tasks with polygons and rectangles.
+        link: https://labelstud.io/guide/export.html#COCO
+        tags:
+        - image segmentation
+        - object detection
+        name: COCO
+      - title: COCO with Images
+        description: COCO format with images downloaded.
+        link: https://labelstud.io/guide/export.html#COCO
+        tags:
+        - image segmentation
+        - object detection
+        name: COCO_WITH_IMAGES
+      - title: YOLO
+        description: Popular TXT format is created for each image file. Each txt file contains
+          annotations for the corresponding image file, that is object class, object coordinates,
+          height & width.
+        link: https://labelstud.io/guide/export.html#YOLO
+        tags:
+        - image segmentation
+        - object detection
+        name: YOLO
+      - title: YOLO with Images
+        description: YOLO format with images downloaded.
+        link: https://labelstud.io/guide/export.html#YOLO
+        tags:
+        - image segmentation
+        - object detection
+        name: YOLO_WITH_IMAGES
+      - title: YOLOv8 OBB
+        description: Popular TXT format is created for each image file. Each txt file contains
+          annotations for the corresponding image file. The YOLO OBB format designates bounding
+          boxes by their four corner points with coordinates normalized between 0 and 1,
+          so it is possible to export rotated objects.
+        link: https://labelstud.io/guide/export.html#YOLO
+        tags:
+        - image segmentation
+        - object detection
+        name: YOLO_OBB
+      - title: YOLOv8 OBB with Images
+        description: YOLOv8 OBB format with images downloaded.
+        link: https://labelstud.io/guide/export.html#YOLO
+        tags:
+        - image segmentation
+        - object detection
+        name: YOLO_OBB_WITH_IMAGES
+      - title: CONLL2003
+        description: Popular format used for the CoNLL-2003 named entity recognition challenge.
+        link: https://labelstud.io/guide/export.html#CONLL2003
+        tags:
+        - sequence labeling
+        - text tagging
+        - named entity recognition
+        name: CONLL2003
         disabled: true
-      - title: 'Pascal VOC XML'
-        description: 'Popular XML format used for object detection and polygon image segmentation tasks.'
-        link: 'https://labelstud.io/guide/export.html#Pascal-VOC-XML'
-        tags: ['image segmentation', 'object detection']
-        name: 'VOC'
+      - title: Pascal VOC XML
+        description: Popular XML format used for object detection and polygon image segmentation
+          tasks.
+        link: https://labelstud.io/guide/export.html#Pascal-VOC-XML
+        tags:
+        - image segmentation
+        - object detection
+        name: VOC
         disabled: true
-      - title: 'Brush labels to NumPy'
-        description: 'Export your brush labels as NumPy 2d arrays. Each label outputs as one image.'
-        link: 'https://labelstud.io/guide/export.html#Brush-labels-to-NumPy-amp-PNG'
-        tags: ['image segmentation']
-        name: 'BRUSH_TO_NUMPY'
+      - title: Brush labels to NumPy
+        description: Export your brush labels as NumPy 2d arrays. Each label outputs as
+          one image.
+        link: https://labelstud.io/guide/export.html#Brush-labels-to-NumPy-amp-PNG
+        tags:
+        - image segmentation
+        name: BRUSH_TO_NUMPY
         disabled: true
-      - title: 'Brush labels to PNG'
-        description: 'Export your brush labels as PNG images. Each label outputs as one image.'
-        link: 'https://labelstud.io/guide/export.html#Brush-labels-to-NumPy-amp-PNG'
-        tags: ['image segmentation']
-        name: 'BRUSH_TO_PNG'
+      - title: Brush labels to PNG
+        description: Export your brush labels as PNG images. Each label outputs as one image.
+        link: https://labelstud.io/guide/export.html#Brush-labels-to-NumPy-amp-PNG
+        tags:
+        - image segmentation
+        name: BRUSH_TO_PNG
         disabled: true
-      - title: 'ASR Manifest'
-        description: 'Export audio transcription labels for automatic speech recognition as the JSON manifest format expected by NVIDIA NeMo models.'
-        link: 'https://labelstud.io/guide/export.html#ASR-MANIFEST'
-        tags: ['speech recognition']
-        name: 'ASR_MANIFEST'
+      - title: ASR Manifest
+        description: Export audio transcription labels for automatic speech recognition
+          as the JSON manifest format expected by NVIDIA NeMo models.
+        link: https://labelstud.io/guide/export.html#ASR-MANIFEST
+        tags:
+        - speech recognition
+        name: ASR_MANIFEST
         disabled: true
+
     status_code: 200