Segment Industrial Defects with Florence-2 and Detectron2

                      
                     
import os
import cv2
import torch
from typing import List, Dict, Any
from detectron2.engine import DefaultPredictor
from detectron2.config import get_cfg
from detectron2 import model_zoo

# Configuration
class Config:
    MODEL_PATH: str = os.getenv('MODEL_PATH', 'path/to/your/model')
    CONFIDENCE_THRESHOLD: float = 0.5

# Initialize the Detectron2 model
def setup_model() -> DefaultPredictor:
    cfg = get_cfg()
    cfg.merge_from_file(model_zoo.get_config_file('COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml'))
    cfg.MODEL.WEIGHTS = Config.MODEL_PATH
    cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = Config.CONFIDENCE_THRESHOLD
    return DefaultPredictor(cfg)

# Function to segment defects in an image
def segment_defects(image_path: str) -> List[Dict[str, Any]]:
    try:
        # Load image
        image = cv2.imread(image_path)
        predictor = setup_model()
        outputs = predictor(image)
        # Process outputs
        return outputs['instances'].to('cpu').get_fields()
    except Exception as e:
        print(f'Error during defect segmentation: {str(e)}')
        return []

if __name__ == '__main__':
    result = segment_defects('path/to/image.jpg')
    print(result)