app.py

import os
from pathlib import Path
from typing import List, Union
from PIL import Image
import ezdxf.units
import numpy as np
import torch
from torchvision import transforms
from ultralytics import YOLOWorld, YOLO
from ultralytics.engine.results import Results
from ultralytics.utils.plotting import save_one_box
from transformers import AutoModelForImageSegmentation
import cv2
import ezdxf
import gradio as gr
import gc
from scalingtestupdated import calculate_scaling_factor
from scipy.interpolate import splprep, splev
from scipy.ndimage import gaussian_filter1d
import json
import time
import signal
from shapely.ops import unary_union
from shapely.geometry import MultiPolygon, GeometryCollection, Polygon, Point
from u2netp import U2NETP  # Add U2NETP import
import logging
import shutil

# Initialize logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

# Create cache directory for models
CACHE_DIR = os.path.join(os.path.dirname(__file__), ".cache")
os.makedirs(CACHE_DIR, exist_ok=True)

# Custom Exception Classes
class TimeoutReachedError(Exception):
    pass

class BoundaryOverlapError(Exception):
    pass

class TextOverlapError(Exception):
    pass

class ReferenceBoxNotDetectedError(Exception):
    """Raised when the Reference coin cannot be detected in the image"""
    pass

class FingerCutOverlapError(Exception):
    """Raised when finger cuts overlap with existing geometry"""
    def __init__(self, message="There was an overlap with fingercuts... Please try again to generate dxf."):
        super().__init__(message)

# Global model initialization
print("Loading models...")
start_time = time.time()

# Load YOLO reference model
reference_model_path = os.path.join("", "best1.pt")
if not os.path.exists(reference_model_path):
    shutil.copy("best1.pt", reference_model_path)
reference_detector_global = YOLO(reference_model_path)

# Load U2NETP model
u2net_model_path = os.path.join(CACHE_DIR, "u2netp.pth")
if not os.path.exists(u2net_model_path):
    shutil.copy("u2netp.pth", u2net_model_path)
u2net_global = U2NETP(3, 1)
u2net_global.load_state_dict(torch.load(u2net_model_path, map_location="cpu"))

# Load BiRefNet model
birefnet = AutoModelForImageSegmentation.from_pretrained(
    "zhengpeng7/BiRefNet", trust_remote_code=True, cache_dir=CACHE_DIR
)

device = "cpu"
torch.set_float32_matmul_precision(["high", "highest"][0])

# Move models to device
u2net_global.to(device)
u2net_global.eval()
birefnet.to(device)
birefnet.eval()

# Define transforms
transform_image = transforms.Compose([
    transforms.Resize((1024, 1024)),
    transforms.ToTensor(),
    transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
])

# Language translations dictionary remains unchanged
TRANSLATIONS = {
    "english": {
        "input_image": "Input Image",
        "offset_value": "Offset value",  
        "offset_unit": "Offset unit (mm/in)",
        "enable_finger": "Enable Finger Clearance",
        "edge_radius": "Edge rounding radius (mm)",
        "output_image": "Output Image",
        "outlines": "Outlines of Objects",
        "dxf_file": "DXF file",
        "mask": "Mask",
        "enable_radius": "Enable Edge Rounding",
        "radius_disabled": "Rounding Disabled",
        "scaling_factor": "Scaling Factor(mm)",
        "scaling_placeholder": "Every pixel is equal to mentioned number in millimeters",
        "language_selector": "Select Language",
    },
    "dutch": {
        "input_image": "Invoer Afbeelding",
        "offset_value": "Offset waarde",
        "offset_unit": "Offset unit (mm/inch)",
        "enable_finger": "Finger Clearance inschakelen",
        "edge_radius": "Ronding radius rand (mm)",
        "output_image": "Uitvoer Afbeelding",
        "outlines": "Contouren van Objecten",
        "dxf_file": "DXF bestand",
        "mask": "Masker",
        "enable_radius": "Ronding inschakelen",
        "radius_disabled": "Ronding uitgeschakeld",
        "scaling_factor": "Schalingsfactor(mm)",
        "scaling_placeholder": "Elke pixel is gelijk aan genoemd aantal in millimeters",
        "language_selector": "Selecteer Taal",
    }
}

def remove_bg_u2netp(image: np.ndarray) -> np.ndarray:
    """Remove background using U2NETP model specifically for reference objects"""
    try:
        image_pil = Image.fromarray(image)
        transform_u2netp = transforms.Compose([
            transforms.Resize((320, 320)),
            transforms.ToTensor(),
            transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
        ])
        
        input_tensor = transform_u2netp(image_pil).unsqueeze(0).to(device)
        
        with torch.no_grad():
            outputs = u2net_global(input_tensor)
        
        pred = outputs[0]
        pred = (pred - pred.min()) / (pred.max() - pred.min() + 1e-8)
        pred_np = pred.squeeze().cpu().numpy()
        pred_np = cv2.resize(pred_np, (image_pil.width, image_pil.height))
        pred_np = (pred_np * 255).astype(np.uint8)
        
        return pred_np
    except Exception as e:
        logger.error(f"Error in U2NETP background removal: {e}")
        raise

def remove_bg(image: np.ndarray) -> np.ndarray:
    """Remove background using BiRefNet model for main objects"""
    try:
        image = Image.fromarray(image)
        input_images = transform_image(image).unsqueeze(0).to(device)

        with torch.no_grad():
            preds = birefnet(input_images)[-1].sigmoid().cpu()
        pred = preds[0].squeeze()

        pred_pil: Image = transforms.ToPILImage()(pred)
        
        scale_ratio = 1024 / max(image.size)
        scaled_size = (int(image.size[0] * scale_ratio), int(image.size[1] * scale_ratio))
        
        return np.array(pred_pil.resize(scaled_size))
    except Exception as e:
        logger.error(f"Error in BiRefNet background removal: {e}")
        raise

def resize_img(img: np.ndarray, resize_dim):
    return np.array(Image.fromarray(img).resize(resize_dim))

def make_square(img: np.ndarray):
    """Make the image square by padding"""
    height, width = img.shape[:2]
    max_dim = max(height, width)
    
    pad_height = (max_dim - height) // 2
    pad_width = (max_dim - width) // 2
    
    pad_height_extra = max_dim - height - 2 * pad_height
    pad_width_extra = max_dim - width - 2 * pad_width
    
    if len(img.shape) == 3:  # Color image
        padded = np.pad(
            img,
            (
                (pad_height, pad_height + pad_height_extra),
                (pad_width, pad_width + pad_width_extra),
                (0, 0),
            ),
            mode="edge",
        )
    else:  # Grayscale image
        padded = np.pad(
            img,
            (
                (pad_height, pad_height + pad_height_extra),
                (pad_width, pad_width + pad_width_extra),
            ),
            mode="edge",
        )
    
    return padded


def detect_reference_square(img) -> tuple:
    """Detect reference square in the image and ignore other coins"""
    try:
        res = reference_detector_global.predict(img, conf=0.75)
        if not res or len(res) == 0 or len(res[0].boxes) == 0:
            raise ReferenceBoxNotDetectedError("Unable to detect the reference coin in the image.")
        
        # Get all detected boxes
        boxes = res[0].cpu().boxes.xyxy
        
        # Find the largest box (most likely the reference coin)
        largest_box = None
        max_area = 0
        for box in boxes:
            x_min, y_min, x_max, y_max = box
            area = (x_max - x_min) * (y_max - y_min)
            if area > max_area:
                max_area = area
                largest_box = box
        
        return (
            save_one_box(largest_box.unsqueeze(0), img, save=False),
            largest_box
        )
    except Exception as e:
        if not isinstance(e, ReferenceBoxNotDetectedError):
            logger.error(f"Error in reference square detection: {e}")
            raise ReferenceBoxNotDetectedError("Error detecting reference coin. Please try again with a clearer image.")
        raise


def exclude_scaling_box(
    image: np.ndarray,
    bbox: np.ndarray,
    orig_size: tuple,
    processed_size: tuple,
    expansion_factor: float = 1.2,
) -> np.ndarray:
    x_min, y_min, x_max, y_max = map(int, bbox)
    scale_x = processed_size[1] / orig_size[1]
    scale_y = processed_size[0] / orig_size[0]
    
    x_min = int(x_min * scale_x)
    x_max = int(x_max * scale_x)
    y_min = int(y_min * scale_y)
    y_max = int(y_max * scale_y)
    
    box_width = x_max - x_min
    box_height = y_max - y_min
    
    expanded_x_min = max(0, int(x_min - (expansion_factor - 1) * box_width / 2))
    expanded_x_max = min(
        image.shape[1], int(x_max + (expansion_factor - 1) * box_width / 2)
    )
    expanded_y_min = max(0, int(y_min - (expansion_factor - 1) * box_height / 2))
    expanded_y_max = min(
        image.shape[0], int(y_max + (expansion_factor - 1) * box_height / 2)
    )
    
    image[expanded_y_min:expanded_y_max, expanded_x_min:expanded_x_max] = 0
    return image





def resample_contour(contour, edge_radius_px: int = 0):
    """Resample contour with radius-aware smoothing and periodic handling."""
    logger.info(f"Starting resample_contour with contour of shape {contour.shape}")

    num_points = 1500
    sigma = max(2, int(edge_radius_px) // 4)  # Adjust sigma based on radius

    if len(contour) < 4:  # Need at least 4 points for spline with periodic condition
        error_msg = f"Contour must have at least 4 points, but has {len(contour)} points."
        logger.error(error_msg)
        raise ValueError(error_msg)

    try:
        contour = contour[:, 0, :]
        logger.debug(f"Reshaped contour to shape {contour.shape}")

        # Ensure contour is closed by making start and end points the same
        if not np.array_equal(contour[0], contour[-1]):
            contour = np.vstack([contour, contour[0]])

        # Create periodic spline representation
        tck, u = splprep(contour.T, u=None, s=0, per=True)
        
        # Evaluate spline at evenly spaced points
        u_new = np.linspace(u.min(), u.max(), num_points)
        x_new, y_new = splev(u_new, tck, der=0)
        
        # Apply Gaussian smoothing with wrap-around
        if sigma > 0:
            x_new = gaussian_filter1d(x_new, sigma=sigma, mode='wrap')
            y_new = gaussian_filter1d(y_new, sigma=sigma, mode='wrap')
            
            # Re-close the contour after smoothing
            x_new[-1] = x_new[0]
            y_new[-1] = y_new[0]

        result = np.array([x_new, y_new]).T
        logger.info(f"Completed resample_contour with result shape {result.shape}")
        return result

    except Exception as e:
        logger.error(f"Error in resample_contour: {e}")
        raise






# def save_dxf_spline(inflated_contours, scaling_factor, height, finger_clearance=False):
#     doc = ezdxf.new(units=ezdxf.units.MM)
#     doc.header["$INSUNITS"] = ezdxf.units.MM
#     msp = doc.modelspace()
#     final_polygons_inch = []
#     finger_centers = []
#     original_polygons = []

#     for contour in inflated_contours:
#         try:
#             # Removed the second parameter since it was causing the error
#             resampled_contour = resample_contour(contour)  
            
#             points_inch = [(x * scaling_factor, (height - y) * scaling_factor) 
#                           for x, y in resampled_contour]
            
#             if len(points_inch) < 3:
#                 continue

#             tool_polygon = build_tool_polygon(points_inch)
#             original_polygons.append(tool_polygon)
            
#             if finger_clearance:
#                 try:
#                     tool_polygon, center = place_finger_cut_adjusted(
#                         tool_polygon, points_inch, finger_centers, final_polygons_inch
#                     )
#                 except FingerCutOverlapError:
#                     tool_polygon = original_polygons[-1]
            
#             exterior_coords = polygon_to_exterior_coords(tool_polygon)
#             if len(exterior_coords) < 3:
#                 continue
            
#             msp.add_spline(exterior_coords, degree=3, dxfattribs={"layer": "TOOLS"})
#             final_polygons_inch.append(tool_polygon)
            
#         except ValueError as e:
#             logger.warning(f"Skipping contour: {e}")

#     dxf_filepath = os.path.join("./outputs", "out.dxf")
#     doc.saveas(dxf_filepath)
#     return dxf_filepath, final_polygons_inch, original_polygons




def save_dxf_spline(inflated_contours, scaling_factor, height, finger_clearance=False):
    doc = ezdxf.new(units=ezdxf.units.MM)
    doc.header["$INSUNITS"] = ezdxf.units.MM
    msp = doc.modelspace()
    final_polygons_inch = []
    finger_centers = []
    original_polygons = []

    # Scale correction factor based on your analysis
    scale_correction = 1.079

    for contour in inflated_contours:
        try:
            resampled_contour = resample_contour(contour)  

            points_inch = [(x * scaling_factor, (height - y) * scaling_factor) 
                          for x, y in resampled_contour]

            if len(points_inch) < 3:
                continue

            tool_polygon = build_tool_polygon(points_inch)
            original_polygons.append(tool_polygon)

            if finger_clearance:
                try:
                    tool_polygon, center = place_finger_cut_adjusted(
                        tool_polygon, points_inch, finger_centers, final_polygons_inch
                    )
                except FingerCutOverlapError:
                    tool_polygon = original_polygons[-1]

            exterior_coords = polygon_to_exterior_coords(tool_polygon)
            if len(exterior_coords) < 3:
                continue

            # Apply scale correction AFTER finger cuts and polygon adjustments
            corrected_coords = [(x * scale_correction, y * scale_correction) for x, y in exterior_coords]

            msp.add_spline(corrected_coords, degree=3, dxfattribs={"layer": "TOOLS"})
            final_polygons_inch.append(tool_polygon)

        except ValueError as e:
            logger.warning(f"Skipping contour: {e}")

    dxf_filepath = os.path.join("./outputs", "out.dxf")
    doc.saveas(dxf_filepath)
    return dxf_filepath, final_polygons_inch, original_polygons





def build_tool_polygon(points_inch):
    return Polygon(points_inch)



def polygon_to_exterior_coords(poly):
    logger.info(f"Starting polygon_to_exterior_coords with input geometry type: {poly.geom_type}")

    try:
        # 1) If it's a GeometryCollection or MultiPolygon, fuse everything into one shape
        if poly.geom_type == "GeometryCollection" or poly.geom_type == "MultiPolygon":
            logger.debug(f"Performing unary_union on {poly.geom_type}")
            unified = unary_union(poly)
            if unified.is_empty:
                logger.warning("unary_union produced an empty geometry; returning empty list")
                return []
            # If union still yields multiple disjoint pieces, pick the largest Polygon
            if unified.geom_type == "GeometryCollection" or unified.geom_type == "MultiPolygon":
                largest = None
                max_area = 0.0
                for g in getattr(unified, "geoms", []):
                    if hasattr(g, "area") and g.area > max_area and hasattr(g, "exterior"):
                        max_area = g.area
                        largest = g
                if largest is None:
                    logger.warning("No valid Polygon found in unified geometry; returning empty list")
                    return []
                poly = largest
            else:
                # Now unified should be a single Polygon or LinearRing
                poly = unified

        # 2) At this point, we must have a single Polygon (or something with an exterior)
        if not hasattr(poly, "exterior") or poly.exterior is None:
            logger.warning("Input geometry has no exterior ring; returning empty list")
            return []

        raw_coords = list(poly.exterior.coords)
        total = len(raw_coords)
        logger.info(f"Extracted {total} raw exterior coordinates")

        if total == 0:
            return []

        # 3) Subsample coordinates to at most 100 points (evenly spaced)
        max_pts = 100
        if total > max_pts:
            step = total // max_pts
            sampled = [raw_coords[i] for i in range(0, total, step)]
            # Ensure we include the last point to close the loop
            if sampled[-1] != raw_coords[-1]:
                sampled.append(raw_coords[-1])
            logger.info(f"Downsampled perimeter from {total} to {len(sampled)} points")
            return sampled
        else:
            return raw_coords

    except Exception as e:
        logger.error(f"Error in polygon_to_exterior_coords: {e}")
        return []








def place_finger_cut_adjusted(
    tool_polygon: Polygon,
    points_inch: list,
    existing_centers: list,
    all_polygons: list,
    circle_diameter: float = 25.4,
    min_gap: float = 0.5,
    max_attempts: int = 100
) -> (Polygon, tuple):
    logger.info(f"Starting place_finger_cut_adjusted with {len(points_inch)} input points")

    from shapely.geometry import Point
    import numpy as np
    import time
    import random

    # Fallback: if we run out of time or attempts, place in the "middle" of the outline
    def fallback_solution():
        logger.warning("Using fallback approach for finger cut placement")
        # Pick the midpoint of the original outline as a last-resort center
        fallback_center = points_inch[len(points_inch) // 2]
        r = circle_diameter / 2.0
        fallback_circle = Point(fallback_center).buffer(r, resolution=32)
        try:
            union_poly = tool_polygon.union(fallback_circle)
        except Exception as e:
            logger.warning(f"Fallback union failed ({e}); trying buffer-union fallback")
            union_poly = tool_polygon.buffer(0).union(fallback_circle.buffer(0))

        existing_centers.append(fallback_center)
        logger.info(f"Fallback finger cut placed at {fallback_center}")
        return union_poly, fallback_center

    # Precompute values
    r = circle_diameter / 2.0
    needed_center_dist = circle_diameter + min_gap

    # 1) Get perimeter coordinates of this polygon
    raw_perimeter = polygon_to_exterior_coords(tool_polygon)
    if not raw_perimeter:
        logger.warning("No valid exterior coords found; using fallback immediately")
        return fallback_solution()

    # 2) Possibly subsample to at most 100 perimeter points
    if len(raw_perimeter) > 100:
        step = len(raw_perimeter) // 100
        perimeter_coords = raw_perimeter[::step]
        logger.info(f"Subsampled perimeter from {len(raw_perimeter)} to {len(perimeter_coords)} points")
    else:
        perimeter_coords = raw_perimeter[:]

    # 3) Randomize the order to avoid bias
    indices = list(range(len(perimeter_coords)))
    random.shuffle(indices)
    logger.debug(f"Shuffled perimeter indices for candidate order")

    # 4) Non-blocking timeout setup
    start_time = time.time()
    timeout_secs = 5.0  # leave ~0.1s margin

    attempts = 0
    try:
        while attempts < max_attempts:
            # 5) Abort if we're running out of time
            if time.time() - start_time > timeout_secs - 0.1:
                logger.warning(f"Approaching timeout after {attempts} attempts")
                return fallback_solution()

            # 6) For each shuffled perimeter point, try small offsets
            for idx in indices:
                # Check timeout inside the loop as well
                if time.time() - start_time > timeout_secs - 0.05:
                    logger.warning("Timeout during candidate-point loop")
                    return fallback_solution()

                cx, cy = perimeter_coords[idx]
                # Try five small offsets: (0,0), (±min_gap/2, 0), (0, ±min_gap/2)
                for dx, dy in [(0, 0), (-min_gap/2, 0), (min_gap/2, 0), (0, -min_gap/2), (0, min_gap/2)]:
                    candidate_center = (cx + dx, cy + dy)

                    # 6a) Check distance to existing finger centers
                    too_close_finger = any(
                        np.hypot(candidate_center[0] - ex, candidate_center[1] - ey) 
                        < needed_center_dist 
                        for (ex, ey) in existing_centers
                    )
                    if too_close_finger:
                        continue

                    # 6b) Build candidate circle with reduced resolution for speed
                    candidate_circle = Point(candidate_center).buffer(r, resolution=32)

                    # 6c) Must overlap ≥30% with this polygon
                    try:
                        inter_area = tool_polygon.intersection(candidate_circle).area
                    except Exception:
                        continue

                    if inter_area < 0.3 * candidate_circle.area:
                        continue

                    # 6d) Must not intersect or even "touch" any other polygon (buffered by min_gap)
                    invalid = False
                    for other_poly in all_polygons:
                        if other_poly.equals(tool_polygon):
                            # Don't compare against itself
                            continue
                        # Buffer the other polygon by min_gap to enforce a strict clearance
                        if other_poly.buffer(min_gap).intersects(candidate_circle) or \
                           other_poly.buffer(min_gap).touches(candidate_circle):
                            invalid = True
                            break
                    if invalid:
                        continue

                    # 6e) Candidate passes all tests → union and return
                    try:
                        union_poly = tool_polygon.union(candidate_circle)
                        # If union is a MultiPolygon (more than one piece), reject
                        if union_poly.geom_type == "MultiPolygon" and len(union_poly.geoms) > 1:
                            continue
                        # If union didn't change anything (no real cut), reject
                        if union_poly.equals(tool_polygon):
                            continue
                    except Exception:
                        continue

                    existing_centers.append(candidate_center)
                    logger.info(f"Finger cut placed successfully at {candidate_center} after {attempts} attempts")
                    return union_poly, candidate_center

            attempts += 1
            # If we've done half the attempts and we're near timeout, bail out
            if attempts >= (max_attempts // 2) and (time.time() - start_time) > timeout_secs * 0.8:
                logger.warning(f"Approaching timeout (attempt {attempts})")
                return fallback_solution()

            logger.debug(f"Completed iteration {attempts}/{max_attempts}")

        # If we exit loop without finding a valid spot
        logger.warning(f"No valid spot after {max_attempts} attempts, using fallback")
        return fallback_solution()

    except Exception as e:
        logger.error(f"Error in place_finger_cut_adjusted: {e}")
        return fallback_solution()










def extract_outlines(binary_image: np.ndarray) -> tuple:
    contours, _ = cv2.findContours(
        binary_image, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE
    )

    outline_image = np.full_like(binary_image, 255)  # White background

    return outline_image, contours




def round_edges(mask: np.ndarray, radius_mm: float, scaling_factor: float) -> np.ndarray:
    """Rounds mask edges using contour smoothing."""
    if radius_mm <= 0 or scaling_factor <= 0:
        return mask
    
    radius_px = max(1, int(radius_mm / scaling_factor))  # Ensure min 1px
    
    # Handle small objects
    if np.count_nonzero(mask) < 500:  # Small object threshold
        return cv2.dilate(cv2.erode(mask, np.ones((3,3))), np.ones((3,3)))
    
    # Existing contour processing with improvements:
    contours, _ = cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
    
    # NEW: Filter small contours
    contours = [c for c in contours if cv2.contourArea(c) > 100]
    smoothed_contours = []
    
    for contour in contours:
        try:
            # Resample with radius-based smoothing
            resampled = resample_contour(contour, radius_px)
            resampled = resampled.astype(np.int32).reshape((-1, 1, 2))
            smoothed_contours.append(resampled)
        except Exception as e:
            logger.warning(f"Error smoothing contour: {e}")
            smoothed_contours.append(contour)  # Fallback to original contour
    
    # Draw smoothed contours
    rounded = np.zeros_like(mask)
    cv2.drawContours(rounded, smoothed_contours, -1, 255, thickness=cv2.FILLED)
    
    return rounded


def predict_og(image, offset, offset_unit, edge_radius, finger_clearance=False):
    print(f"DEBUG: Image shape: {image.shape}, dtype: {image.dtype}, range: {image.min()}-{image.max()}")

    coin_size_mm = 20.0

    if offset_unit == "inches":
        offset *= 25.4

    if edge_radius is None or edge_radius == 0:
        edge_radius = 0.0001

    if offset < 0:
        raise gr.Error("Offset Value Can't be negative")

    try:
        reference_obj_img, scaling_box_coords = detect_reference_square(image)
    except ReferenceBoxNotDetectedError as e:
        return (
        None,
        None,
        None,
        None,
        f"Error: {str(e)}"
    )
    except Exception as e:
        raise gr.Error(f"Error processing image: {str(e)}")

    reference_obj_img = make_square(reference_obj_img)
    
    # Use U2NETP for reference object background removal
    reference_square_mask = remove_bg_u2netp(reference_obj_img)
    reference_square_mask = resize_img(reference_square_mask, reference_obj_img.shape[:2][::-1])

    try:
        scaling_factor = calculate_scaling_factor(
            target_image=reference_square_mask,
            reference_obj_size_mm=coin_size_mm,
            feature_detector="ORB",
        )
    except Exception as e:
        scaling_factor = None
        logger.warning(f"Error calculating scaling factor: {e}")

    if not scaling_factor:
        ref_size_px = (reference_square_mask.shape[0] + reference_square_mask.shape[1]) / 2
        scaling_factor = 20.0 / ref_size_px
        logger.info(f"Fallback scaling: {scaling_factor:.4f} mm/px using 20mm reference")

    # Use BiRefNet for main object background removal
    orig_size = image.shape[:2]
    objects_mask = remove_bg(image)
    processed_size = objects_mask.shape[:2]

    # REMOVE ALL COINS from mask:
    res = reference_detector_global.predict(image, conf=0.05)
    boxes = res[0].cpu().boxes.xyxy if res and len(res) > 0 else []

    for box in boxes:
        objects_mask = exclude_scaling_box(
            objects_mask,
            box,
            orig_size,
            processed_size,
            expansion_factor=1.2,
        )

    objects_mask = resize_img(objects_mask, (image.shape[1], image.shape[0]))

    # offset_pixels = (float(offset) / scaling_factor) * 2 + 1 if scaling_factor else 1
    # dilated_mask = cv2.dilate(objects_mask, np.ones((int(offset_pixels), int(offset_pixels)), np.uint8))
    # Image.fromarray(dilated_mask).save("./outputs/scaled_mask_original.jpg")
    # dilated_mask_orig = dilated_mask.copy()

    # #if edge_radius > 0:
    #     # Use morphological rounding instead of contour-based
    # rounded_mask = round_edges(objects_mask, edge_radius, scaling_factor)
    # #else:
    #     #rounded_mask = objects_mask.copy()
    
    # # Apply dilation AFTER rounding
    # offset_pixels = (float(offset) / scaling_factor) * 2 + 1 if scaling_factor else 1
    # kernel = np.ones((int(offset_pixels), int(offset_pixels)), np.uint8)
    # dilated_mask = cv2.dilate(rounded_mask, kernel)
    # Apply edge rounding first
    rounded_mask = round_edges(objects_mask, edge_radius, scaling_factor)

    # Apply dilation AFTER rounding
    offset_pixels = (float(offset) / scaling_factor) * 2 + 1 if scaling_factor else 1
    kernel = np.ones((int(offset_pixels), int(offset_pixels)), np.uint8)
    final_dilated_mask = cv2.dilate(rounded_mask, kernel)

    # Save for debugging
    Image.fromarray(final_dilated_mask).save("./outputs/scaled_mask_original.jpg")
        

    outlines, contours = extract_outlines(final_dilated_mask)

    try:
        dxf, finger_polygons, original_polygons = save_dxf_spline(
            contours,
            scaling_factor,
            processed_size[0],
            finger_clearance=(finger_clearance == "On")
        )
    except FingerCutOverlapError as e:
        raise gr.Error(str(e))

    shrunked_img_contours = image.copy()

    if finger_clearance == "On":
        outlines = np.full_like(final_dilated_mask, 255)
        for poly in finger_polygons:
            try:
                coords = np.array([
                    (int(x / scaling_factor), int(processed_size[0] - y / scaling_factor))
                    for x, y in poly.exterior.coords
                ], np.int32).reshape((-1, 1, 2))

                cv2.drawContours(shrunked_img_contours, [coords], -1, 0, thickness=2)
                cv2.drawContours(outlines, [coords], -1, 0, thickness=2)
            except Exception as e:
                logger.warning(f"Failed to draw finger cut: {e}")
                continue
    else:
        outlines = np.full_like(final_dilated_mask, 255)
        cv2.drawContours(shrunked_img_contours, contours, -1, 0, thickness=2)
        cv2.drawContours(outlines, contours, -1, 0, thickness=2)

    return (
    shrunked_img_contours,
    outlines,
    dxf,
    final_dilated_mask,
    f"{scaling_factor:.4f}")


def predict_simple(image):
    """
    Only image in → returns (annotated, outlines, dxf, mask).
    Uses offset=0 mm, no fillet, no finger-cut.
    """
    ann, outlines, dxf_path, mask, _ = predict_og(
        image,
        offset=0,
        offset_unit="mm",
        edge_radius=0,
        finger_clearance="Off",
    )
    return ann, outlines, dxf_path, mask

def predict_middle(image, enable_fillet, fillet_value_mm):
    """
    image + (On/Off) fillet toggle + fillet radius → returns (annotated, outlines, dxf, mask).
    Uses offset=0 mm, finger-cut off.
    """
    radius = fillet_value_mm if enable_fillet == "On" else 0
    ann, outlines, dxf_path, mask, _ = predict_og(
        image,
        offset=0,
        offset_unit="mm",
        edge_radius=radius,
        finger_clearance="Off",
    )
    return ann, outlines, dxf_path, mask

def predict_full(image, enable_fillet, fillet_value_mm, enable_finger_cut):
    """
    image + fillet toggle/value + finger-cut toggle → returns (annotated, outlines, dxf, mask).
    Uses offset=0 mm.
    """
    radius = fillet_value_mm if enable_fillet == "On" else 0
    finger_flag = "On" if enable_finger_cut == "On" else "Off"
    ann, outlines, dxf_path, mask, _ = predict_og(
        image,
        offset=0,
        offset_unit="mm",
        edge_radius=radius,
        finger_clearance=finger_flag,
    )
    return ann, outlines, dxf_path, mask




def update_interface(language):
    return [
        gr.Image(label=TRANSLATIONS[language]["input_image"], type="numpy"),
        gr.Row([
            gr.Number(label=TRANSLATIONS[language]["offset_value"], value=0),
            gr.Dropdown(["mm", "inches"], value="mm", 
                      label=TRANSLATIONS[language]["offset_unit"])
        ]),
        gr.Slider(minimum=0,maximum=20,step=1,value=5,label=TRANSLATIONS[language]["edge_radius"],visible=False,interactive=True),
        gr.Radio(choices=["On", "Off"],value="Off",label=TRANSLATIONS[language]["enable_radius"],),
        gr.Image(label=TRANSLATIONS[language]["output_image"]),
        gr.Image(label=TRANSLATIONS[language]["outlines"]),
        gr.File(label=TRANSLATIONS[language]["dxf_file"]),
        gr.Image(label=TRANSLATIONS[language]["mask"]),
        gr.Textbox(label=TRANSLATIONS[language]["scaling_factor"],placeholder=TRANSLATIONS[language]["scaling_placeholder"],),
    ]

if __name__ == "__main__":
    os.makedirs("./outputs", exist_ok=True)

    with gr.Blocks() as demo:
        language = gr.Dropdown(
            choices=["english", "dutch"],
            value="english",
            label="Select Language",
            interactive=True
        )
        
        input_image = gr.Image(label=TRANSLATIONS["english"]["input_image"], type="numpy")
        
        with gr.Row():
            offset = gr.Number(label=TRANSLATIONS["english"]["offset_value"], value=0)
            offset_unit = gr.Dropdown([
                "mm", "inches"
            ], value="mm", label=TRANSLATIONS["english"]["offset_unit"])
            
        finger_toggle = gr.Radio(
            choices=["On", "Off"],
            value="Off",
            label=TRANSLATIONS["english"]["enable_finger"]
        )
        
        edge_radius = gr.Slider(
            minimum=0,
            maximum=20,
            step=1,
            value=5,
            label=TRANSLATIONS["english"]["edge_radius"],
            visible=False,
            interactive=True
        )
        
        radius_toggle = gr.Radio(
            choices=["On", "Off"],
            value="Off",
            label=TRANSLATIONS["english"]["enable_radius"],
            interactive=True
        )
        
        def toggle_radius(choice):
            if choice == "On":
                return gr.Slider(visible=True)
            return gr.Slider(visible=False, value=0)

        radius_toggle.change(
            fn=toggle_radius,
            inputs=radius_toggle,
            outputs=edge_radius
        )
        
        output_image = gr.Image(label=TRANSLATIONS["english"]["output_image"])
        outlines = gr.Image(label=TRANSLATIONS["english"]["outlines"])
        dxf_file = gr.File(label=TRANSLATIONS["english"]["dxf_file"])
        mask = gr.Image(label=TRANSLATIONS["english"]["mask"])
        
        scaling = gr.Textbox(
            label=TRANSLATIONS["english"]["scaling_factor"],
            placeholder=TRANSLATIONS["english"]["scaling_placeholder"]
        )

        submit_btn = gr.Button("Submit")
        
        language.change(
            fn=lambda x: [
                gr.update(label=TRANSLATIONS[x]["input_image"]),
                gr.update(label=TRANSLATIONS[x]["offset_value"]),
                gr.update(label=TRANSLATIONS[x]["offset_unit"]),
                gr.update(label=TRANSLATIONS[x]["output_image"]),
                gr.update(label=TRANSLATIONS[x]["outlines"]),
                gr.update(label=TRANSLATIONS[x]["enable_finger"]),
                gr.update(label=TRANSLATIONS[x]["dxf_file"]),
                gr.update(label=TRANSLATIONS[x]["mask"]),
                gr.update(label=TRANSLATIONS[x]["enable_radius"]),
                gr.update(label=TRANSLATIONS[x]["edge_radius"]),
                gr.update(
                    label=TRANSLATIONS[x]["scaling_factor"],
                    placeholder=TRANSLATIONS[x]["scaling_placeholder"]
                ),
            ],
            inputs=[language],
            outputs=[
                input_image, offset, offset_unit,
                output_image, outlines, finger_toggle, dxf_file,
                mask, radius_toggle, edge_radius, scaling
            ]
        )
        
        def custom_predict_and_format(*args):
            output_image, outlines, dxf_path, mask, scaling = predict_og(*args)
            if output_image is None:
                return (
                    None, None, None, None, "Reference coin not detected!"
                )
            return (
                output_image, outlines, dxf_path, mask, scaling
            )

        submit_btn.click(
            fn=custom_predict_and_format,
            inputs=[input_image, offset, offset_unit, edge_radius, finger_toggle],
            outputs=[output_image, outlines, dxf_file, mask, scaling]
        )


        gr.Examples(
            examples=[
                ["./examples/Test20.jpg", 0, "mm"],
                ["./examples/Test21.jpg", 0, "mm"],
                ["./examples/Test22.jpg", 0, "mm"],
                ["./examples/Test23.jpg", 0, "mm"],
            ],
            inputs=[input_image, offset, offset_unit]
        )

    demo.launch(share=True)