Update handler.py

handler.py

@@ -1,26 +1,32 @@
-from typing import Dict, List, Any
+from transformers import pipeline
 from PIL import Image
 from io import BytesIO
-from transformers import pipeline
 import base64
+from typing import Dict, List, Any
 
 class EndpointHandler():
-    def __init__(self, path=""):
-        self.pipeline = pipeline(task="zero-shot-object-detection",model=path, device = 0 ) #device = 0 to use GPU rather than -1 which would be CPU
+    def __init__(self, model_path=""):
+        # Initialize the pipeline with the specified model and set the device to GPU
+        self.pipeline = pipeline(task="zero-shot-object-detection", model=model_path, device=0)
 
     def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
         """
-       data args:
-            images (:obj:`string`)
-            candiates (:obj:`list`)
-      Return:
-            A :obj:`list`:. The list contains items that are dicts should be liked {"label": "XXX", "score": 0.82}
+        Process an incoming request for zero-shot object detection.
+        
+        Args:
+            data (Dict[str, Any]): The input data containing an encoded image and candidate labels.
+            
+        Returns:
+            A list of dictionaries, each containing a label and its corresponding score.
         """
-        inputs = data.pop("inputs", data)
-
-        # decode base64 image to PIL
+        # Correctly accessing the 'inputs' key and fixing the typo in 'candidates'
+        inputs = data.get("inputs", {})
+        
+        # Decode the base64 image to a PIL image
         image = Image.open(BytesIO(base64.b64decode(inputs['image'])))
 
-        # run prediction one image wit provided candiates
-        detector = self.pipeline(images=[image], candidate_labels=inputs["candidate_labels"])
-        return detector
+        # Correctly passing the image and candidate labels to the pipeline
+        detection_results = self.pipeline(image=image, candidate_labels=inputs["candidates"])
+        
+        # Adjusting the return statement to match the expected output structure
+        return detection_results