Example for object detection with ultralytics YOLO v8 model

examples/object_detector/yolo/yolov8/README.md

@@ -0,0 +1,62 @@
+# Object Detection using Ultralytics's pretrained YOLOv8(yolov8n) model.
+
+
+Install `ultralytics` using
+```
+python -m pip install -r requirements.txt
+```
+
+In this example, we are using the YOLOv8 Nano model from ultralytics.Downlaod the pretrained weights from [Ultralytics](https://docs.ultralytics.com/models/yolov8/#supported-modes)
+
+```
+wget https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8n.pt
+```
+
+## Create a model archive file for Yolov8n model
+
+```
+torch-model-archiver --model-name yolov8n --version 1.0 --serialized-file yolov8n.pt --handler custom_handler.py
+```
+
+```
+mkdir model_store
+mv yolov8n.mar model_store/.
+```
+
+## Start TorchServe and register the model
+
+
+```
+torchserve --start --model-store model_store --ncs
+curl -X POST "localhost:8081/models?model_name=yolov8n&url=yolov8n.mar&initial_workers=4&batch_size=2"
+```
+
+results in
+
+```
+{
+  "status": "Model \"yolov8n\" Version: 1.0 registered with 4 initial workers"
+}
+```
+
+## Run Inference
+
+Here we are counting the number of detected objects in the image. You can change the post-process method in the handler to return the bounding box coordinates
+
+```
+curl http://127.0.0.1:8080/predictions/yolov8n -T persons.jpg  & curl http://127.0.0.1:8080/predictions/yolov8n -T bus.jpg
+```
+
+gives the output
+
+```
+{
+  "person": 4,
+  "handbag": 3,
+  "bench": 3
+}{
+  "person": 4,
+  "bus": 1,
+  "stop sign": 1
+}
+```

examples/object_detector/yolo/yolov8/custom_handler.py

@@ -0,0 +1,88 @@
+import logging
+import os
+from collections import defaultdict
+
+import torch
+from torchvision import transforms
+from ultralytics import YOLO
+
+from ts.torch_handler.object_detector import ObjectDetector
+from ts.utils.util import load_label_mapping
+
+logger = logging.getLogger(__name__)
+
+try:
+    import torch_xla.core.xla_model as xm
+
+    XLA_AVAILABLE = True
+except ImportError as error:
+    XLA_AVAILABLE = False
+
+
+class Yolov8Handler(ObjectDetector):
+    image_processing = transforms.Compose(
+        [transforms.Resize(640), transforms.CenterCrop(640), transforms.ToTensor()]
+    )
+
+    def __init__(self):
+        super(Yolov8Handler, self).__init__()
+
+    def initialize(self, context):
+        properties = context.system_properties
+        if torch.cuda.is_available() and properties.get("gpu_id") is not None:
+            self.map_location = "cuda"
+            self.device = torch.device(
+                self.map_location + ":" + str(properties.get("gpu_id"))
+            )
+        elif XLA_AVAILABLE:
+            self.device = xm.xla_device()
+        else:
+            self.map_location = "cpu"
+            self.device = torch.device(self.map_location)
+
+        self.manifest = context.manifest
+
+        model_dir = properties.get("model_dir")
+        self.model_pt_path = None
+        if "serializedFile" in self.manifest["model"]:
+            serialized_file = self.manifest["model"]["serializedFile"]
+            self.model_pt_path = os.path.join(model_dir, serialized_file)
+
+        self.model = self._load_torchscript_model(self.model_pt_path)
+
+        logger.debug("Model file %s loaded successfully", self.model_pt_path)
+
+        # Load class mapping for classifiers
+        mapping_file_path = os.path.join(model_dir, "index_to_name.json")
+        self.mapping = load_label_mapping(mapping_file_path)
+
+        self.initialized = True
+
+    def _load_torchscript_model(self, model_pt_path):
+        """Loads the PyTorch model and returns the NN model object.
+
+        Args:
+            model_pt_path (str): denotes the path of the model file.
+
+        Returns:
+            (NN Model Object) : Loads the model object.
+        """
+        # TODO: remove this method if https://github.com/pytorch/text/issues/1793 gets resolved
+
+        model = YOLO(model_pt_path)
+        model.to(self.device)
+        return model
+
+    def postprocess(self, res):
+        output = []
+        for data in res:
+            classes = data.boxes.cls.tolist()
+            names = data.names
+
+            result = defaultdict(int)
+
+            for cls in classes:
+                result[names[int(cls)]] += 1
+            output.append(result)
+
+        return output

examples/object_detector/yolo/yolov8/requirements.txt

@@ -0,0 +1 @@
+ultralytics>=8.0.144