app.py

import gradio as gr
import torch
import cv2
import numpy as np
from torchvision import transforms
from PIL import Image

# Load MiDaS depth estimation model
midas_model = torch.hub.load("intel-isl/MiDaS", "DPT_Hybrid")
midas_model.eval()
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
midas_model.to(device)
midas_transform = torch.hub.load("intel-isl/MiDaS", "transforms").default_transform

def estimate_depth(image):
    """Estimate depth map to identify fabric folds."""
    image = image.convert("RGB")
    image_tensor = midas_transform(image).to(device)
    
    with torch.no_grad():
        depth = midas_model(image_tensor).squeeze().cpu().numpy()
    
    depth = cv2.resize(depth, (image.size[0], image.size[1]))
    depth = (depth - depth.min()) / (depth.max() - depth.min()) * 255
    return depth.astype(np.uint8)

def detect_folds(image):
    """Apply edge detection and highlight cloth folds."""
    depth_map = estimate_depth(image)
    edges = cv2.Canny(depth_map, 50, 150)
    
    # Convert edges to 3-channel image for visualization
    edges_colored = cv2.cvtColor(edges, cv2.COLOR_GRAY2RGB)
    overlay = cv2.addWeighted(np.array(image), 0.7, edges_colored, 0.3, 0)
    
    return Image.fromarray(overlay)

def main(image):
    return detect_folds(image)

iface = gr.Interface(
    fn=main,
    inputs=gr.Image(type="pil"),
    outputs=gr.Image(type="pil"),
    title="Cloth Fold Detection",
    description="Upload an image of clothing to visualize folds using depth estimation and edge detection."
)

if __name__ == "__main__":
    iface.launch(share=True, debug=True)