handler.py

from transformers import pipeline
from PIL import Image
from io import BytesIO
import base64
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
        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.
        
        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.
        """
        # 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'])))

        # 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