Upload folder using huggingface_hub

Micro-OD.py

@@ -0,0 +1,120 @@
+"""Micro-OD: A few-shot microscopy object detection benchmark."""
+
+import json
+import os
+
+import datasets
+
+_DESCRIPTION = """\
+Micro-OD is a few-shot microscopy object detection benchmark spanning four
+biological imaging domains: BBBC (malaria parasite detection), BCCD (blood
+cell counting), LIVECell (RatC6 live-cell imaging), and NIH-3T3 (mouse
+fibroblast imaging).
+
+Two splits are provided:
+  - example: 10 images per sub-dataset (40 total) — the few-shot support set.
+  - test:    53 images per sub-dataset (212 total) — the evaluation query set.
+"""
+
+# Sorted alphabetically; used as ClassLabel names.
+ALL_CLASSES = [
+    "Gametocyte Cells",
+    "Platelets",
+    "Polygonal Cells",
+    "Red Blood Cells",
+    "Ring Cells",
+    "Round Cells",
+    "Schizont Cells",
+    "Spindle Cells",
+    "Trophozoite Cells",
+    "White Blood Cells",
+]
+
+SUBDATASETS = ["BBBC", "BCCD", "LIVECell", "NIH-3T3"]
+
+
+class MicroOD(datasets.GeneratorBasedBuilder):
+    """Micro-OD dataset loader."""
+
+    VERSION = datasets.Version("1.0.0")
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=datasets.Features(
+                {
+                    "image": datasets.Image(),
+                    "image_id": datasets.Value("string"),
+                    "subdataset": datasets.Value("string"),
+                    "objects": datasets.Sequence(
+                        feature={
+                            # COCO format: [x_min, y_min, width, height]
+                            "bbox": datasets.Sequence(
+                                datasets.Value("float32"), length=4
+                            ),
+                            "label": datasets.ClassLabel(names=ALL_CLASSES),
+                            "category": datasets.Value("string"),
+                        }
+                    ),
+                }
+            ),
+        )
+
+    def _split_generators(self, dl_manager):
+        data_dir = os.path.dirname(os.path.abspath(__file__))
+        return [
+            datasets.SplitGenerator(
+                name=datasets.splits.NamedSplit("example"),
+                gen_kwargs={"split_path": os.path.join(data_dir, "example")},
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.TEST,
+                gen_kwargs={"split_path": os.path.join(data_dir, "test")},
+            ),
+        ]
+
+    def _generate_examples(self, split_path):
+        idx = 0
+        for subdataset in SUBDATASETS:
+            annotation_path = os.path.join(
+                split_path, subdataset, "annotation.jsonl"
+            )
+            image_base = os.path.join(split_path, subdataset)
+
+            with open(annotation_path, encoding="utf-8") as f:
+                for line in f:
+                    line = line.strip()
+                    if not line:
+                        continue
+
+                    entry = json.loads(line)
+                    image_path = os.path.join(image_base, entry["image_path"])
+
+                    bboxes, labels, categories = [], [], []
+                    for category, boxes in entry["bbox"].items():
+                        for box in boxes:
+                            x_min, y_min = box[0]
+                            x_max, y_max = box[1]
+                            # Convert [[x_min,y_min],[x_max,y_max]] → COCO [x,y,w,h]
+                            bboxes.append(
+                                [
+                                    float(x_min),
+                                    float(y_min),
+                                    float(x_max - x_min),
+                                    float(y_max - y_min),
+                                ]
+                            )
+                            labels.append(ALL_CLASSES.index(category))
+                            categories.append(category)
+
+                    yield idx, {
+                        "image": image_path,
+                        "image_id": f"{subdataset}/{entry['image_path']}",
+                        "subdataset": subdataset,
+                        "objects": {
+                            "bbox": bboxes,
+                            "label": labels,
+                            "category": categories,
+                        },
+                    }
+                    idx += 1

README.md

@@ -133,6 +133,29 @@ Annotations are stored as [JSON Lines](https://jsonlines.org/) (`.jsonl`) files
 }
 ```
 
+## Usage
+
+```python
+from datasets import load_dataset
+
+ds = load_dataset("path/to/Micro-OD")
+
+# Access splits
+example_split = ds["example"]   # 40 images — few-shot support set
+test_split    = ds["test"]       # 212 images — evaluation query set
+
+# Each row contains:
+#   image       — PIL image
+#   image_id    — "<subdataset>/images/<filename>"
+#   subdataset  — one of: BBBC, BCCD, LIVECell, NIH-3T3
+#   objects     — dict with keys:
+#                   bbox     : list of [x_min, y_min, width, height]  (COCO format, float)
+#                   label    : list of int  (index into the 10-class label set)
+#                   category : list of str  (class name)
+```
+
+> **Note on bbox format:** The loading script converts the raw `[[x_min, y_min], [x_max, y_max]]` annotation format into COCO-style `[x_min, y_min, width, height]` float lists. If you consume the `annotation.jsonl` files directly, refer to the [Annotation Format](#annotation-format) section for the raw coordinate convention.
+
 ## Dataset Statistics
 
 Detailed per-class statistics are available in `example/stat.txt` and `test/stat.txt`. Summaries are provided below.