File size: 4,786 Bytes
f53b39e |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 |
#!/usr/bin/env python3
"""Sample script to run DepthPro.
Copyright (C) 2024 Apple Inc. All Rights Reserved.
"""
import argparse
import logging
from pathlib import Path
import numpy as np
import PIL.Image
import torch
from matplotlib import pyplot as plt
from tqdm import tqdm
from depth_pro import create_model_and_transforms, load_rgb
LOGGER = logging.getLogger(__name__)
def get_torch_device() -> torch.device:
"""Get the Torch device."""
device = torch.device("cpu")
if torch.cuda.is_available():
device = torch.device("cuda:0")
elif torch.backends.mps.is_available():
device = torch.device("mps")
return device
def run(args):
"""Run Depth Pro on a sample image."""
if args.verbose:
logging.basicConfig(level=logging.INFO)
# Load model.
model, transform = create_model_and_transforms(
device=get_torch_device(),
precision=torch.half,
)
model.eval()
image_paths = [args.image_path]
if args.image_path.is_dir():
image_paths = args.image_path.glob("**/*")
relative_path = args.image_path
else:
relative_path = args.image_path.parent
if not args.skip_display:
plt.ion()
fig = plt.figure()
ax_rgb = fig.add_subplot(121)
ax_disp = fig.add_subplot(122)
for image_path in tqdm(image_paths):
# Load image and focal length from exif info (if found.).
try:
LOGGER.info(f"Loading image {image_path} ...")
image, _, f_px = load_rgb(image_path)
except Exception as e:
LOGGER.error(str(e))
continue
# Run prediction. If `f_px` is provided, it is used to estimate the final metric depth,
# otherwise the model estimates `f_px` to compute the depth metricness.
prediction = model.infer(transform(image), f_px=f_px)
# Extract the depth and focal length.
depth = prediction["depth"].detach().cpu().numpy().squeeze()
if f_px is not None:
LOGGER.debug(f"Focal length (from exif): {f_px:0.2f}")
elif prediction["focallength_px"] is not None:
focallength_px = prediction["focallength_px"].detach().cpu().item()
LOGGER.info(f"Estimated focal length: {focallength_px}")
inverse_depth = 1 / depth
# Visualize inverse depth instead of depth, clipped to [0.1m;250m] range for better visualization.
max_invdepth_vizu = min(inverse_depth.max(), 1 / 0.1)
min_invdepth_vizu = max(1 / 250, inverse_depth.min())
inverse_depth_normalized = (inverse_depth - min_invdepth_vizu) / (
max_invdepth_vizu - min_invdepth_vizu
)
# Save Depth as npz file.
if args.output_path is not None:
output_file = (
args.output_path
/ image_path.relative_to(relative_path).parent
/ image_path.stem
)
LOGGER.info(f"Saving depth map to: {str(output_file)}")
output_file.parent.mkdir(parents=True, exist_ok=True)
np.savez_compressed(output_file, depth=depth)
# Save as color-mapped "turbo" jpg image.
cmap = plt.get_cmap("turbo")
color_depth = (cmap(inverse_depth_normalized)[..., :3] * 255).astype(
np.uint8
)
color_map_output_file = str(output_file) + ".jpg"
LOGGER.info(f"Saving color-mapped depth to: : {color_map_output_file}")
PIL.Image.fromarray(color_depth).save(
color_map_output_file, format="JPEG", quality=90
)
# Display the image and estimated depth map.
if not args.skip_display:
ax_rgb.imshow(image)
ax_disp.imshow(inverse_depth_normalized, cmap="turbo")
fig.canvas.draw()
fig.canvas.flush_events()
LOGGER.info("Done predicting depth!")
if not args.skip_display:
plt.show(block=True)
def main():
"""Run DepthPro inference example."""
parser = argparse.ArgumentParser(
description="Inference scripts of DepthPro with PyTorch models."
)
parser.add_argument(
"-i",
"--image-path",
type=Path,
default="./data/example.jpg",
help="Path to input image.",
)
parser.add_argument(
"-o",
"--output-path",
type=Path,
help="Path to store output files.",
)
parser.add_argument(
"--skip-display",
action="store_true",
help="Skip matplotlib display.",
)
parser.add_argument(
"-v",
"--verbose",
action="store_true",
help="Show verbose output."
)
run(parser.parse_args())
if __name__ == "__main__":
main()
|