Update handler.py

handler.py

@@ -6,30 +6,30 @@ from typing import Dict, List, Any
 
 class EndpointHandler():
     def __init__(self, model_path=""):
-        # Initialize the pipeline with the specified model and set the device to GPU
+        # Initialize the zero-shot object detection pipeline with the specified model
+        # and set the device to GPU for faster computation.
         self.pipeline = pipeline(task="zero-shot-object-detection", model=model_path, device=0)
 
     def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
         """
-        Process an incoming request for zero-shot object detection.
+        Handles incoming requests for zero-shot object detection, decoding the image
+        and predicting labels based on provided candidates.
         
         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.
+            List[Dict[str, Any]]: Predictions with labels and scores for the detected objects.
         """
-        # 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'])))
+        # Decode the base64-encoded image to a PIL Image object for processing.
+        image_data = data.get("inputs", {}).get('image', '')
+        image = Image.open(BytesIO(base64.b64decode(image_data)))
 
-        # Get candidate labels
-        candidate_labels=inputs["candidates"]
+        # Extract candidate labels from the input data.
+        candidate_labels = data.get("inputs", {}).get("candidates", [])
 
-        # Correctly passing the image and candidate labels to the pipeline
-        detection_results = self.pipeline(image=image, candidate_labels=inputs["candidates"])
+        # Perform zero-shot object detection using the provided image and candidate labels.
+        detection_results = self.pipeline(image=image, candidate_labels=candidate_labels)
         
-        # Adjusting the return statement to match the expected output structure
-        return detection_results
+        # Return the detection results directly, which should match the expected output structure.
+        return detection_results