handler.py

# handler.py

import io
import base64
import numpy as np
from PIL import Image
import torch
from transformers import SamModel, SamProcessor
from typing import Dict, List, Any
import torch.nn.functional as F

# set device
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

class EndpointHandler():
    def __init__(self, path=""):
        """
        Called once at startup. 
        Load the SAM model using Hugging Face Transformers.
        """
        try:
            # Load the model and processor from the local path
            self.model = SamModel.from_pretrained(path).to(device)
            self.processor = SamProcessor.from_pretrained(path)
        except Exception as e:
            # Fallback to loading from a known SAM model if local loading fails
            print("Failed to load from local path: {}".format(e))
            print("Attempting to load from facebook/sam-vit-base")
            self.model = SamModel.from_pretrained("facebook/sam-vit-base").to(device)
            self.processor = SamProcessor.from_pretrained("facebook/sam-vit-base")

    def __call__(self, data):
        """
        Called on every HTTP request.
        Handles both base64-encoded images and PIL images.
        Returns a PIL Image object.
        """
        # 1. Parse and decode the input image
        inputs = data.pop("inputs", None)
        if inputs is None:
            raise ValueError("Missing 'inputs' key in the payload.")

        # Check the type of inputs to handle both base64 strings and pre-processed PIL Images
        if isinstance(inputs, Image.Image):
            img = inputs.convert("RGB")
        elif isinstance(inputs, str):
            if inputs.startswith("data:"):
                inputs = inputs.split(",", 1)[1]
            image_bytes = base64.b64decode(inputs)
            img = Image.open(io.BytesIO(image_bytes)).convert("RGB")
        else:
            raise TypeError("Unsupported input type. Expected a PIL Image or a base64 encoded string.")
        
        # 2. Prepare prompts and process the image
        height, width = img.size[1], img.size[0]
        input_points = [[[width // 2, height // 2]]]
        input_labels = [[1]]
        
        inputs = self.processor(img, input_points=input_points, input_labels=input_labels, return_tensors="pt").to(device)
        
        # 3. Generate masks
        with torch.no_grad():
            outputs = self.model(**inputs)
        
        # 4. Process and select the best mask
        try:
            original_height, original_width = inputs["original_sizes"][0].tolist()
            pred_masks = outputs.pred_masks.cpu()
            iou_scores = outputs.iou_scores.cpu()[0]

            if pred_masks.ndim == 5:
                pred_masks = pred_masks.squeeze(1)

            best_mask_idx = torch.argmax(iou_scores)
            best_mask_tensor = pred_masks[0, best_mask_idx, :, :]

            upscaled_mask = F.interpolate(
                best_mask_tensor.unsqueeze(0).unsqueeze(0).float(),
                size=(original_height, original_width),
                mode='bilinear',
                align_corners=False
            ).squeeze()

            mask_binary = (upscaled_mask > 0.0).numpy().astype(np.uint8) * 255
        
        except Exception as e:
            print("Error processing masks: {}".format(e))
            mask_binary = np.zeros((height, width), dtype=np.uint8)
            center_x, center_y = width // 2, height // 2
            size = min(width, height) // 8
            mask_binary[center_y-size:center_y+size, center_x-size:center_x+size] = 255
        
        # 5. Create and return the output PIL Image
        output_img = Image.fromarray(mask_binary)
        return output_img

def main():
    # This main function shows how a client would call the endpoint locally.
    input_path = "/Users/rp7/Downloads/test.jpeg"
    output_path = "output.png"

    # 1. Prepare the payload with a base64-encoded image string
    with open(input_path, "rb") as f:
        img_bytes = f.read()
    img_b64 = base64.b64encode(img_bytes).decode("utf-8")
    payload = {"inputs": "data:image/jpeg;base64,{}".format(img_b64)}

    # 2. Instantiate handler and get the PIL Image result
    handler = EndpointHandler(path=".")
    result_img = handler(payload)

    # 3. Save the resulting image
    result_img.save(output_path)
    print("Wrote mask to {}".format(output_path))

if __name__ == "__main__":
    main()