SiPEED-A075V/depth.py

import cv2
import torch
import time
import numpy as np
# import open3d as o3d

# Load MiDas model for depth estimation
model_type = "DPT_Hybrid"
midas = torch.hub.load("intel-isl/MiDas", model_type)
device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
midas.to(device)
midas.eval()

# Load transformers to resize and normalize the image
midas_transforms = torch.hub.load("intel-isl/MiDas", "transforms")
transform = midas_transforms.dpt_transform if model_type == "DPT_Large" or model_type == "DPT_Hybrid" else midas_transforms.small_transform


# Create Open3D visualizer
# vis = o3d.visualization.Visualizer()
# vis.create_window("3D Point Cloud")

# Create a geometry object to store the point cloud
# pcd = o3d.geometry.PointCloud()

def img2depth(img_rgb):
    # Apply input transform
    input_batch = transform(img_rgb).to(device)

    # Prediction and resize to original resolution
    with torch.no_grad():
        prediction = midas(input_batch)
        prediction = torch.nn.functional.interpolate(
            prediction.unsqueeze(1),
            size=img_rgb.shape[:2],
            mode="bicubic",
            align_corners=False,
        ).squeeze()

    return prediction.cpu().numpy()

    # Apply colormap to depth map
    # depth_colored = cv2.applyColorMap((depth_map / depth_map.max() * 255).astype(np.uint8), cv2.COLORMAP_JET)

    # Generate 3D point cloud with colors
    # point_cloud = []
    # colors = []
    # for y in range(depth_map.shape[0]):
    #     for x in range(depth_map.shape[1]):
    #         depth = depth_map[y, x]
    #         if depth > 0:
    #             # Convert pixel coordinates to 3D space
    #             z = depth
    #             x_3d = x
    #             y_3d = depth_map.shape[0] - y
    #             # point_cloud.append([x_3d, y_3d, z])
    #             colors.append(img_rgb[y, x])

    # point_cloud = np.array(point_cloud)
    # colors = np.array(colors)

    # Update point cloud data
    # pcd.points = o3d.utility.Vector3dVector(point_cloud)
    # pcd.colors = o3d.utility.Vector3dVector(colors / 255.0)

    # Scale point cloud to increase its size
    # pcd.scale(1000.0, center=pcd.get_center())

    # Visualize point cloud in 3D space
    # vis.clear_geometries()
    # vis.add_geometry(pcd)
    # vis.poll_events()
    # vis.update_renderer()

    # Display camera feed and depth map