import torch
import torchvision.transforms as transforms
from PIL import Image
import numpy as np

# Example: Using a pretrained depth estimation model
# (MiDaS is a popular choice for turning 2D images into depth maps)

midas = torch.hub.load("intel-isl/MiDaS", "DPT_Hybrid")
midas.eval()

transform = torch.hub.load("intel-isl/MiDaS", "transforms").dpt_transform

def estimate_depth(image_path):
    img = Image.open(image_path)
    input_batch = transform(img).unsqueeze(0)

    with torch.no_grad():
        prediction = midas(input_batch)
        depth_map = torch.nn.functional.interpolate(
            prediction.unsqueeze(1),
            size=img.size[::-1],
            mode="bicubic",
            align_corners=False,
        ).squeeze()

    return depth_map.cpu().numpy()

# Example usage
depth = estimate_depth("your_picture.jpg")

# Convert depth map into a 3D point cloud
import open3d as o3d

def depth_to_pointcloud(depth_map, scale=1.0):
    h, w = depth_map.shape
    xx, yy = np.meshgrid(np.arange(w), np.arange(h))
    points = np.stack((xx, yy, depth_map * scale), axis=-1).reshape(-1, 3)

    pcd = o3d.geometry.PointCloud()
    pcd.points = o3d.utility.Vector3dVector(points)
    return pcd

pcd = depth_to_pointcloud(depth)
o3d.visualization.draw_geometries([pcd])