Merge branch 'main' into mmdet_tensorrt_support

.github/workflows/ci.yml

@@ -116,10 +116,6 @@ jobs:
         run: >
           pip install ultralytics==8.0.207
 
-      - name: Install super-gradients(3.3.1)
-        run: >
-          pip install super-gradients==3.3.1
-
       - name: Unittest for SAHI+YOLOV5/MMDET/Detectron2 on all platforms
         run: |
           python -m unittest

.github/workflows/package_testing.yml

@@ -93,10 +93,6 @@ jobs:
         run: >
           pip install ultralytics==8.0.207
 
-      - name: Install super-gradients(3.3.1)
-        run: >
-          pip install super-gradients==3.3.1
-
       - name: Install latest SAHI package
         run: >
           pip install --upgrade --force-reinstall sahi

README.md

@@ -34,17 +34,17 @@ Object detection and instance segmentation are by far the most important applica
 
 | Command  | Description  |
 |---|---|
-| [predict](https://github.com/obss/sahi/blob/main/docs/cli.md#predict-command-usage)  | perform sliced/standard video/image prediction using any [yolov5](https://github.com/ultralytics/yolov5)/[mmdet](https://github.com/open-mmlab/mmdetection)/[detectron2](https://github.com/facebookresearch/detectron2)/[huggingface](https://huggingface.co/models?pipeline_tag=object-detection&sort=downloads) model |
-| [predict-fiftyone](https://github.com/obss/sahi/blob/main/docs/cli.md#predict-fiftyone-command-usage)  | perform sliced/standard prediction using any [yolov5](https://github.com/ultralytics/yolov5)/[mmdet](https://github.com/open-mmlab/mmdetection)/[detectron2](https://github.com/facebookresearch/detectron2)/[huggingface](https://huggingface.co/models?pipeline_tag=object-detection&sort=downloads) model and explore results in [fiftyone app](https://github.com/voxel51/fiftyone) |
+| [predict](https://github.com/obss/sahi/blob/main/docs/cli.md#predict-command-usage)  | perform sliced/standard video/image prediction using any [ultralytics](https://github.com/ultralytics/ultralytics)/[mmdet](https://github.com/open-mmlab/mmdetection)/[detectron2](https://github.com/facebookresearch/detectron2)/[huggingface](https://huggingface.co/models?pipeline_tag=object-detection&sort=downloads)/[torchvision](https://pytorch.org/vision/stable/models.html#object-detection) model |
+| [predict-fiftyone](https://github.com/obss/sahi/blob/main/docs/cli.md#predict-fiftyone-command-usage)  | perform sliced/standard prediction using any [ultralytics](https://github.com/ultralytics/ultralytics)/[mmdet](https://github.com/open-mmlab/mmdetection)/[detectron2](https://github.com/facebookresearch/detectron2)/[huggingface](https://huggingface.co/models?pipeline_tag=object-detection&sort=downloads)/[torchvision](https://pytorch.org/vision/stable/models.html#object-detection) model and explore results in [fiftyone app](https://github.com/voxel51/fiftyone) |
 | [coco slice](https://github.com/obss/sahi/blob/main/docs/cli.md#coco-slice-command-usage)  | automatically slice COCO annotation and image files |
 | [coco fiftyone](https://github.com/obss/sahi/blob/main/docs/cli.md#coco-fiftyone-command-usage)  | explore multiple prediction results on your COCO dataset with [fiftyone ui](https://github.com/voxel51/fiftyone) ordered by number of misdetections |
 | [coco evaluate](https://github.com/obss/sahi/blob/main/docs/cli.md#coco-evaluate-command-usage)  | evaluate classwise COCO AP and AR for given predictions and ground truth |
 | [coco analyse](https://github.com/obss/sahi/blob/main/docs/cli.md#coco-analyse-command-usage)  | calculate and export many error analysis plots |
-| [coco yolov5](https://github.com/obss/sahi/blob/main/docs/cli.md#coco-yolov5-command-usage)  | automatically convert any COCO dataset to [yolov5](https://github.com/ultralytics/yolov5) format |
+| [coco yolov5](https://github.com/obss/sahi/blob/main/docs/cli.md#coco-yolov5-command-usage)  | automatically convert any COCO dataset to [ultralytics](https://github.com/ultralytics/ultralytics) format |
 
 ## <div align="center">Quick Start Examples</div>
 
-[📜 List of publications that cite SAHI (currently 150+)](https://scholar.google.com/scholar?hl=en&as_sdt=2005&sciodt=0,5&cites=14065474760484865747&scipsc=&q=&scisbd=1)
+[📜 List of publications that cite SAHI (currently 200+)](https://scholar.google.com/scholar?hl=en&as_sdt=2005&sciodt=0,5&cites=14065474760484865747&scipsc=&q=&scisbd=1)
 
 [🏆 List of competition winners that used SAHI](https://github.com/obss/sahi/discussions/688)
 
@@ -80,9 +80,7 @@ Object detection and instance segmentation are by far the most important applica
 
 - `RT-DETR` + `SAHI` walkthrough: <a href="https://colab.research.google.com/github/obss/sahi/blob/main/demo/inference_for_rtdetr.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="sahi-rtdetr"></a> (NEW)
 
-- `YOLOv8` + `SAHI` walkthrough: <a href="https://colab.research.google.com/github/obss/sahi/blob/main/demo/inference_for_yolov8.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="sahi-yolov8"></a> (NEW)
-
-- `SuperGradients/YOLONAS` + `SAHI`: <a href="https://colab.research.google.com/github/obss/sahi/blob/main/demo/inference_for_yolonas.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="sahi-yolonas"></a> (NEW)
+- `YOLOv8` + `SAHI` walkthrough: <a href="https://colab.research.google.com/github/obss/sahi/blob/main/demo/inference_for_yolov8.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="sahi-yolov8"></a>
 
 - `DeepSparse` + `SAHI` walkthrough: <a href="https://colab.research.google.com/github/obss/sahi/blob/main/demo/inference_for_sparse_yolov5.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="sahi-deepsparse"></a>
 

sahi/models/detectron2.py

@@ -149,9 +149,9 @@ def _create_object_prediction_list_from_original_predictions(
             object_prediction_list = [
                 ObjectPrediction(
                     bbox=box.tolist() if mask is None else None,
-                    segmentation=get_coco_segmentation_from_bool_mask(mask.detach().cpu().numpy())
-                    if mask is not None
-                    else None,
+                    segmentation=(
+                        get_coco_segmentation_from_bool_mask(mask.detach().cpu().numpy()) if mask is not None else None
+                    ),
                     category_id=category_id.item(),
                     category_name=self.category_mapping[str(category_id.item())],
                     shift_amount=shift_amount,

sahi/utils/coco.py

@@ -224,28 +224,7 @@ def __init__(
         self._shapely_annotation = shapely_annotation
 
     def get_sliced_coco_annotation(self, slice_bbox: List[int]):
-        def filter_polygons(geometry):
-            """
-            Filters out and returns only Polygon or MultiPolygon components of a geometry.
-            If geometry is a Polygon, it converts it into a MultiPolygon.
-            If it's a GeometryCollection, it filters to create a MultiPolygon from any Polygons in the collection.
-            Returns an empty MultiPolygon if no Polygon or MultiPolygon components are found.
-            """
-            if isinstance(geometry, Polygon):
-                return MultiPolygon([geometry])
-            elif isinstance(geometry, MultiPolygon):
-                return geometry
-            elif isinstance(geometry, GeometryCollection):
-                polygons = [geom for geom in geometry.geoms if isinstance(geom, Polygon)]
-                return MultiPolygon(polygons) if polygons else MultiPolygon()
-            return MultiPolygon()
-
         shapely_polygon = box(slice_bbox[0], slice_bbox[1], slice_bbox[2], slice_bbox[3])
-        samp = self._shapely_annotation.multipolygon
-        if not samp.is_valid:
-            valid = make_valid(samp)
-            valid = filter_polygons(valid)
-            self._shapely_annotation.multipolygon = valid
         intersection_shapely_annotation = self._shapely_annotation.get_intersection(shapely_polygon)
         return CocoAnnotation.from_shapely_annotation(
             intersection_shapely_annotation,

sahi/utils/shapely.py

@@ -3,7 +3,8 @@
 
 from typing import List
 
-from shapely.geometry import CAP_STYLE, JOIN_STYLE, MultiPolygon, Polygon, box
+from shapely.geometry import CAP_STYLE, JOIN_STYLE, GeometryCollection, MultiPolygon, Polygon, box
+from shapely.validation import make_valid
 
 
 def get_shapely_box(x: int, y: int, width: int, height: int) -> Polygon:
@@ -21,15 +22,36 @@ def get_shapely_box(x: int, y: int, width: int, height: int) -> Polygon:
 
 def get_shapely_multipolygon(coco_segmentation: List[List]) -> MultiPolygon:
     """
-    Accepts coco style polygon coords and converts it to shapely multipolygon object
+    Accepts coco style polygon coords and converts it to valid shapely multipolygon object
     """
+
+    def filter_polygons(geometry):
+        """
+        Filters out and returns only Polygon or MultiPolygon components of a geometry.
+        If geometry is a Polygon, it converts it into a MultiPolygon.
+        If it's a GeometryCollection, it filters
+        to create a MultiPolygon from any Polygons in the collection.
+        Returns an empty MultiPolygon if no Polygon or MultiPolygon components are found.
+        """
+        if isinstance(geometry, Polygon):
+            return MultiPolygon([geometry])
+        elif isinstance(geometry, MultiPolygon):
+            return geometry
+        elif isinstance(geometry, GeometryCollection):
+            polygons = [geom for geom in geometry.geoms if isinstance(geom, Polygon)]
+            return MultiPolygon(polygons) if polygons else MultiPolygon()
+        return MultiPolygon()
+
     polygon_list = []
     for coco_polygon in coco_segmentation:
         point_list = list(zip(coco_polygon[0::2], coco_polygon[1::2]))
         shapely_polygon = Polygon(point_list)
         polygon_list.append(shapely_polygon)
     shapely_multipolygon = MultiPolygon(polygon_list)
 
+    if not shapely_multipolygon.is_valid:
+        shapely_multipolygon = filter_polygons(make_valid(shapely_multipolygon))
+
     return shapely_multipolygon
 
 

tests/test_shapelyutils.py

@@ -26,6 +26,12 @@ def test_get_shapely_multipolygon(self):
         self.assertEqual(shapely_multipolygon.area, 41177.5)
         self.assertTupleEqual(shapely_multipolygon.bounds, (1, 1, 325, 200))
 
+    def test_get_shapely_multipolygon_naughty(self):
+        # self-intersection case
+        coco_segmentation = [[3559.0, 2046.86, 3.49, 2060.0, 3540.9, 3249.7, 2060.0, 3239.61, 2052.87]]
+        shapely_multipolygon = get_shapely_multipolygon(coco_segmentation)
+        self.assertTrue(shapely_multipolygon.is_valid)
+
     def test_shapely_annotation(self):
         # init shapely_annotation from coco segmentation
         segmentation = [[1, 1, 325, 125.2, 250, 200, 5, 200]]