add support for global csv annotation files

Author: tonylampada

roboflow/__init__.py

@@ -14,7 +14,7 @@
 from roboflow.models import CLIPModel, GazeModel  # noqa: F401
 from roboflow.util.general import write_line
 
-__version__ = "1.1.26"
+__version__ = "1.1.27"
 
 
 def check_key(api_key, model, notebook, num_retries=0):

roboflow/core/project.py

@@ -528,9 +528,9 @@ def single_upload(
 
     def _annotation_params(self, annotation_path):
         annotation_name, annotation_string = None, None
-        if isinstance(annotation_path, dict):
+        if isinstance(annotation_path, dict) and annotation_path.get("rawText"):
             annotation_name = annotation_path["name"]
-            annotation_string = json.dumps(annotation_path["parsed"])
+            annotation_string = annotation_path["rawText"]
         elif os.path.exists(annotation_path):
             with open(annotation_path, "r"):
                 annotation_string = open(annotation_path, "r").read()

roboflow/core/workspace.py

@@ -341,8 +341,8 @@ def _upload_image(imagedesc):
             labelmap = None
             annotationdesc = imagedesc.get("annotationfile")
             if annotationdesc:
-                if annotationdesc.get("parsed"):
-                    annotation_path = {"name": annotationdesc["name"], "parsed": annotationdesc["parsed"]}
+                if annotationdesc.get("rawText"):
+                    annotation_path = annotationdesc
                 else:
                     annotation_path = f"{location}{annotationdesc['file']}"
                 labelmap = annotationdesc.get("labelmap")

roboflow/util/folderparser.py

@@ -5,7 +5,7 @@
 from .image_utils import load_labelmap
 
 IMAGE_EXTENSIONS = {".jpg", ".jpeg", ".png", ".bmp"}
-ANNOTATION_EXTENSIONS = {".txt", ".json", ".xml"}
+ANNOTATION_EXTENSIONS = {".txt", ".json", ".xml", ".csv"}
 LABELMAPS_EXTENSIONS = {".labels", ".yaml", ".yml"}
 
 
@@ -138,6 +138,7 @@ def _filterIndividualAnnotations(image, annotation, format):
                     or [fake_annotation],
                 },
             }
+            _annotation["rawText"] = json.dumps(_annotation["parsed"])
             return _annotation
     elif format == "createml":
         imgReferences = [i for i in parsed if i["image"] == image["name"]]
@@ -149,25 +150,58 @@ def _filterIndividualAnnotations(image, annotation, format):
                 "name": "annotation.createml.json",
                 "parsedType": "createml",
                 "parsed": [imgReference],
+                "rawText": json.dumps([imgReference]),
             }
             return _annotation
+    elif format == "csv":
+        imgLines = [l["line"] for l in parsed["lines"] if l["file_name"] == image["name"]]
+        if imgLines:
+            headers = parsed["headers"]
+            _annotation = {
+                "name": "annotation.csv",
+                "parsedType": "csv",
+                "parsed": {
+                    "headers": headers,
+                    "lines": imgLines,
+                },
+                "rawText": "".join([headers] + imgLines),
+            }
+            return _annotation
+        else:
+            return None
     return None
 
 
 def _loadAnnotations(folder, annotations):
-    valid_extensions = {".json"}
+    valid_extensions = {".json", ".csv"}
     annotations = [a for a in annotations if a["extension"] in valid_extensions]
     for ann in annotations:
         extension = ann["extension"]
-        with open(f"{folder}{ann['file']}", "r") as f:
-            parsed = json.load(f)
-            parsedType = _guessAnnotationFileFormat(parsed, extension)
-            if parsedType:
-                ann["parsed"] = parsed
-                ann["parsedType"] = parsedType
+        if extension == ".json":
+            with open(f"{folder}{ann['file']}", "r") as f:
+                parsed = json.load(f)
+                parsedType = _guessAnnotationFileFormat(parsed, extension)
+                if parsedType:
+                    ann["parsed"] = parsed
+                    ann["parsedType"] = parsedType
+        elif extension == ".csv":
+            ann["parsedType"] = "csv"
+            ann["parsed"] = _parseAnnotationCSV(f"{folder}{ann['file']}")
     return annotations
 
 
+def _parseAnnotationCSV(filename):
+    # TODO: use a proper CSV library?
+    with open(filename, "r") as f:
+        lines = f.readlines()
+    headers = lines[0]
+    lines = [{"file_name": l.split(",")[0].strip(), "line": l} for l in lines[1:]]
+    return {
+        "headers": headers,
+        "lines": lines,
+    }
+
+
 def _guessAnnotationFileFormat(parsed, extension):
     if extension == ".json":
         if isinstance(parsed, dict):

tests/datasets/mosquitos/mosquitos.csv

@@ -0,0 +1,5 @@
+img_fName,img_w,img_h,class_label,bbx_xtl,bbx_ytl,bbx_xbr,bbx_ybr
+train_10307.jpeg,3024,4032,culex,1459,1389,1826,2062
+train_10308.jpeg,1058,943,japonicus/koreicus,28,187,908,815
+train_10309.jpeg,1024,1365,culex,304,438,614,785
+train_10310.jpeg,2976,3968,albopictus,1900,1280,2163,1653

tests/datasets/mosquitos/train_10307.jpeg

@@ -40,5 +40,6 @@
         # f"import {thisdir}/data/cultura-pepino-yolov8 -w wolfodorpythontests -p yellow-auto -c 100".split()  # noqa: E501 // docs
         # f"import {thisdir}/data/cultura-pepino-yolov8_voc -w wolfodorpythontests -p yellow-auto -c 100".split()  # noqa: E501 // docs
         f"import {thisdir}/data/cultura-pepino-yolov5pytorch -w wolfodorpythontests -p yellow-auto -c 100 -n papaiasso".split()  # noqa: E501 // docs
+        # f"import {thisdir}/../datasets/mosquitos -w wolfodorpythontests -p yellow-auto -n papaiasso".split()  # noqa: E501 // docs
     )
     args.func(args)

tests/datasets/mosquitos/train_10308.jpeg

@@ -38,6 +38,17 @@ def test_parse_sharks_yolov9(self):
         assert testImage["annotationfile"]["file"] == expectAnnotationFile
         assert testImage["annotationfile"]["labelmap"] == {0: "fish", 1: "primary", 2: "shark"}
 
+    def test_parse_mosquitos_csv(self):
+        sharksfolder = f"{thisdir}/../datasets/mosquitos"
+        parsed = folderparser.parsefolder(sharksfolder)
+        testImagePath = "/train_10308.jpeg"
+        testImage = [i for i in parsed["images"] if i["file"] == testImagePath][0]
+        assert testImage["annotationfile"]["name"] == "annotation.csv"
+        headers = testImage["annotationfile"]["parsed"]["headers"]
+        lines = testImage["annotationfile"]["parsed"]["lines"]
+        assert headers == "img_fName,img_w,img_h,class_label,bbx_xtl,bbx_ytl,bbx_xbr,bbx_ybr\n"
+        assert lines == ["train_10308.jpeg,1058,943,japonicus/koreicus,28,187,908,815\n"]
+
 
 def _assertJsonMatchesFile(actual, filename):
     with open(filename, "r") as file: