File size: 5,154 Bytes
b35f113 | 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 | """
Pooling Gain — Automatum Data Highway Area Pools (Multi-Drone)
Compares a pooled (two-drone) recording with its two single-drone recordings:
observed road extent, track length distribution and share of tracks > 1 km.
Usage:
python 04_pooling_gain.py <path_to_pool_folder> [<single_drone_1_folder> <single_drone_2_folder>] [--plot out.png]
If the single-drone folders are not given, the script looks for them next to the
pool folder (same location name, e.g. Highway-A9-OffenbauRastanlage_01_... and _02_...).
This is the layout after extracting automatum_data_highway_area_pools.zip.
If they are not found (e.g. when only Sample_Data/ was downloaded), only the
pool itself is analysed.
Example:
python 04_pooling_gain.py ../Sample_Data/Pool-Highway-A9-OffenbauRastanlage_c025-c02598c7-fa60-4302-8cee-12a05f4e0db6
python 04_pooling_gain.py ../automatum_data_highway_area_pools/Pool-Highway-A9-OffenbauRastanlage_c025-c02598c7-fa60-4302-8cee-12a05f4e0db6 --plot pooling_gain.png
Definitions (same as in the dataset card):
road extent = 0.5-99.5 % extent of all vehicle positions along the main road axis
track length = distance driven per vehicle while it is observed
"""
import sys
import os
import re
import numpy as np
from openautomatumdronedata.dataset import droneDataset
def track_stats(path):
dataset = droneDataset(path)
objs = [o for o in dataset.dynWorld.get_list_of_dynamic_objects() if len(o.x_vec) > 1]
lengths = np.array([np.sum(np.hypot(np.diff(o.x_vec), np.diff(o.y_vec))) for o in objs])
pts = np.vstack([np.column_stack([o.x_vec, o.y_vec]) for o in objs])
c = pts.mean(0)
_, _, vt = np.linalg.svd(pts[::20] - c, full_matrices=False)
along = (pts - c) @ vt[0]
lo, hi = np.percentile(along, [0.5, 99.5])
return {
"name": os.path.basename(os.path.normpath(path)),
"n": len(objs),
"road_extent_m": float(hi - lo),
"median_len_m": float(np.median(lengths)),
"p90_len_m": float(np.percentile(lengths, 90)),
"share_gt_1km": float(np.mean(lengths > 1000.0)),
"vehicle_km": float(lengths.sum() / 1000.0),
"lengths": lengths,
}
def find_singles(pool_path):
pool_name = os.path.basename(os.path.normpath(pool_path))
m = re.match(r"Pool-(.+?)_[0-9a-f]{4}-", pool_name)
if not m:
return []
location = m.group(1)
parent = os.path.dirname(os.path.normpath(pool_path))
cands = sorted((d for d in os.listdir(parent)
if d.startswith(location + "_") and not d.startswith("Pool-")
and os.path.isfile(os.path.join(parent, d, "dynamicWorld.json"))),
key=lambda d: ("_02_" in d, d)) # drone 1 first, drone 2 ("_02_") second
return [os.path.join(parent, d) for d in cands]
def main(argv):
plot_file = None
if "--plot" in argv:
i = argv.index("--plot")
plot_file = argv[i + 1]
argv = argv[:i] + argv[i + 2:]
pool = argv[0]
singles = argv[1:3] if len(argv) >= 3 else find_singles(pool)
print(f"Loading pool: {pool}")
rows = [track_stats(pool)]
if not singles:
print("Single-drone recordings not found next to the pool folder -> analysing the pool only.\n"
"Extract automatum_data_highway_area_pools.zip to compare pool vs. single drones.")
for s in singles:
print(f"Loading single-drone recording: {s}")
rows.append(track_stats(s))
print("\n{:<28} {:>7} {:>11} {:>11} {:>10} {:>9} {:>10}".format(
"item", "tracks", "extent [m]", "median [m]", "p90 [m]", "> 1 km", "veh-km"))
for k, r in enumerate(rows):
label = "pool" if k == 0 else f"single drone {k}"
print("{:<28} {:>7d} {:>11.0f} {:>11.0f} {:>10.0f} {:>8.1f}% {:>10.0f}".format(
label, r["n"], r["road_extent_m"], r["median_len_m"], r["p90_len_m"],
100 * r["share_gt_1km"], r["vehicle_km"]))
if len(rows) > 1:
best = max(rows[1:], key=lambda r: r["road_extent_m"])
print(f"\nPooling gain: road extent x{rows[0]['road_extent_m'] / best['road_extent_m']:.2f} "
f"vs. best single drone, median track length x{rows[0]['median_len_m'] / best['median_len_m']:.2f}, "
f"tracks > 1 km: {100 * best['share_gt_1km']:.1f}% -> {100 * rows[0]['share_gt_1km']:.1f}%")
if plot_file:
import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt
bins = np.arange(0, 1500, 50)
plt.figure(figsize=(8, 4.5))
for k, r in enumerate(rows):
label = "pool" if k == 0 else f"single drone {k}"
plt.hist(r["lengths"], bins=bins, histtype="step", lw=2 if k == 0 else 1.2, label=label)
plt.axvline(1000, color="k", ls="--", lw=0.8)
plt.xlabel("observed track length [m]"); plt.ylabel("vehicles"); plt.legend()
plt.title(rows[0]["name"], fontsize=8)
plt.tight_layout(); plt.savefig(plot_file, dpi=120)
print(f"Histogram saved: {os.path.abspath(plot_file)}")
if __name__ == "__main__":
if len(sys.argv) < 2:
print(__doc__)
else:
main(sys.argv[1:])
|