def run():
opentraj_root = sys.argv[1]
output_dir = sys.argv[2]
datasets = get_datasets(opentraj_root, all_dataset_names)
for ds in datasets:
grouping(ds)
def load_dataset(opentraj_root, dataset_name):
if os.path.exists(opentraj_root) \
and os.path.exists(os.path.join(opentraj_root, 'datasets')) and dataset_name:
global traj_dataset, traj_fig, cur_time
cur_time = 0
traj_datasets = get_datasets(opentraj_root, [dataset_name])
if dataset_name in traj_datasets:
traj_dataset = traj_datasets[dataset_name].data
return np.max(traj_dataset["timestamp"])
return 0
sns.swarmplot(y='dca', x='title', data=df_dca, size=1, ax=ax2)
ax2.set_xlabel('')
ax2.set_yticks([0, 1.0, 2.0, 3.0, 4.0, 5.0])
ax2.xaxis.set_tick_params(labelsize=8)
ax2.yaxis.label.set_size(8)
ax2.yaxis.set_tick_params(labelsize=8)
sns.swarmplot(y='collision_energy', x='title', data=df_E, size=1, ax=ax3)
ax3.set_xlabel('')
plt.xticks(rotation=-20)
ax3.set_yticks([0, 4, 8, 12])
ax3.xaxis.set_tick_params(labelsize=8)
ax3.yaxis.label.set_size(8)
ax3.yaxis.set_tick_params(labelsize=8)
plt.subplots_adjust(hspace=0.1)
fig.align_ylabels()
plt.savefig(os.path.join(output_dir, 'collision.pdf'),
dpi=500,
bbox_inches='tight')
plt.show()
if __name__ == "__main__":
opentraj_root = sys.argv[1]
output_dir = sys.argv[2]
datasets = get_datasets(opentraj_root, all_dataset_names)
run(datasets, output_dir)
pcf_accum += pcf_values_t
avg_pcf = pcf_accum / len(frames)
plt.title(dataset.title)
plt.ylabel('PCF')
plt.xlabel('meter')
plt.plot(avg_pcf, color='purple')
# Fixme: the program needs the two following inputs:
if __name__ == '__main__':
opentraj_root = sys.argv[
1] # e.g. os.path.expanduser("~") + '/workspace2/OpenTraj'
output_dir = sys.argv[
2] # e.g. os.path.expanduser("~") + '/Dropbox/OpenTraj-paper/exp/ver-0.2'
datasets = get_datasets(opentraj_root)
# ................................................................
# ================================================================
# =================== List the metric functions ==================
# ================================================================
metrics = [
speed_plot,
# acceleration_plot,
# path_efficiency_plot,
# density_vanilla_plot,
# min_dist_plot,
# pcf_plot
]
# ================================================================
import pandas as pd
import numpy as np
import os
import sys
import matplotlib.pyplot as plt
import plotly.express as px
import plotly.graph_objs as go
from toolkit.benchmarking.load_all_datasets import get_datasets, get_trajlets, all_dataset_names
from toolkit.benchmarking.indicators.motion_properties import speed_avg_of_trajs, acceleration_of_tarjs
from matplotlib import cm
if __name__ == '__main__':
opentraj_root = sys.argv[1]
datasets = get_datasets(opentraj_root, ['KITTI'])
trajlet_sets = get_trajlets(
opentraj_root,
# all_dataset_names[:1]
[
# 'ETH-Univ',
# 'ETH-Hotel',
# 'UCY-Zara',
# 'UCY-Univ',
# 'SDD-coupa',
# 'SDD-bookstore',
# 'SDD-deathCircle',
# 'WildTrack',
# 'KITTI'
'LCas-Minerva',
])
def main():
opentraj_root = sys.argv[1]
output_dir = sys.argv[2]
all_names = [
'ETH-Univ', 'ETH-Hotel', 'UCY-Zara', 'UCY-Univ', 'SDD-Coupa',
'SDD-bookstore', 'SDD-deathCircle', 'GC', 'InD-1', 'InD-2', 'KITTI',
'LCas-Minerva', 'WildTrack', 'Edinburgh', 'BN-1d-w180', 'BN-2d-w160'
]
datasets = get_datasets(opentraj_root, all_names)
#store all the results in pandas dataframe
all_global_density = []
all_local_density = []
# Get trajectories from dataset
for name in all_names:
dataset = datasets[name]
all_frames = dataset.get_frames()
all_trajs = dataset.get_trajectories()
trajlets = {}
#get scene area
scenes_maxX = dataset.data.groupby(['scene_id'])['pos_x'].max()
scenes_minX = dataset.data.groupby(['scene_id'])['pos_x'].min()
scenes_maxY = dataset.data.groupby(['scene_id'])['pos_y'].max()
scenes_minY = dataset.data.groupby(['scene_id'])['pos_y'].min()
area = pd.DataFrame(data=[], columns=['scene_id', 'area'])
for idx in scenes_maxX.index:
x_range = scenes_maxX.loc[idx] - scenes_minX.loc[idx]
y_range = scenes_maxY.loc[idx] - scenes_minY.loc[idx]
area.loc[idx, 'area'] = x_range * y_range
#calculate and store global density
global_dens = global_density(all_frames, area)
g_density = pd.DataFrame(data=np.zeros((len(global_dens), 2)),
columns=['name', 'global_density'])
g_density.iloc[:, 0] = [name for i in range(len(global_dens))]
g_density.iloc[:, 1] = global_dens
all_global_density.append(global_dens)
outputFile1 = output_dir + "/" + name + '_globalDens.h5'
fw = open(outputFile1, 'wb')
pickle.dump(g_density, fw)
fw.close()
#calculate and store local density
trajlets = {}
local_dens = local_density(all_frames, trajlets, name)
l_density = pd.DataFrame(data=[], columns=['name', 'local_density'])
l_density.iloc[:, 1] = local_dens
l_density.iloc[:, 0] = [name for i in range(len(l_density.iloc[:, 1]))]
all_local_density.append(local_dens)
outputFile2 = output_dir + "/" + name + '_localDens.h5'
fw = open(outputFile2, 'wb')
pickle.dump(l_density, fw)
fw.close()
print(name, " finish")
# down-sample each group.
gdens_d = normalize_samples_with_histogram(all_global_density,
max_n_samples=800,
n_bins=50,
quantile_interval=[0.05, 0.98])
ldens_d = normalize_samples_with_histogram(all_local_density,
max_n_samples=800,
n_bins=50,
quantile_interval=[0.05, 0.95])
# put samples in a DataFrame (required for seaborn plots)
df_gdens = pd.concat([
pd.DataFrame({
'title': all_names[ii],
'global_density': gdens_d[ii],
}) for ii in range(len(all_names[:-2]))
])
df_ldens = pd.concat([
pd.DataFrame({
'title': all_names[ii],
'local_density': ldens_d[ii],
}) for ii in range(len(all_names[:-2]))
])
BN_gdens = pd.concat([
pd.DataFrame({
'title': all_names[ii],
'global_density': gdens_d[ii],
}) for ii in range(len(all_names[-2:]))
])
BN_ldens = pd.concat([
pd.DataFrame({
'title': all_names[ii],
'local_density': ldens_d[ii],
}) for ii in range(len(all_names[-2:]))
])
print("making plots ...")
sns.set(style="whitegrid")
fig, axs = plt.subplots(2,
2,
figsize=(12, 2),
gridspec_kw={
'width_ratios': [4, 1],
'height_ratios': [1, 1]
})
sns.swarmplot(y='global_density',
x='title',
data=df_gdens,
size=1,
ax=axs[0, 0])
axs[0, 0].set_ylim([0, 0.08])
axs[0, 0].set_yticks([0, 0.02, 0.04, 0.06, 0.08])
axs[0, 0].set_xlabel('')
axs[0, 0].yaxis.label.set_size(9)
axs[0, 0].yaxis.set_tick_params(labelsize=8)
sns.swarmplot(y='local_density',
x='title',
data=df_ldens,
size=1,
ax=axs[1, 0])
axs[1, 0].set_ylim([0, 6])
axs[1, 0].set_yticks([0, 2.0, 4.0, 6.0])
axs[1, 0].yaxis.label.set_size(9)
axs[1, 0].xaxis.set_tick_params(labelsize=8)
axs[1, 0].set_xlabel('')
axs[1, 0].tick_params(axis='x', labelrotation=-20)
axs[1, 0].yaxis.set_tick_params(labelsize=8)
sns.swarmplot(y='global_density',
x='title',
data=BN_gdens,
size=1,
ax=axs[0, 1])
axs[0, 1].set_ylim([0, 1.5])
axs[0, 1].set_yticks([0, 0.5, 1, 1.5])
axs[0, 1].set_xlabel('')
axs[0, 1].yaxis.label.set_size(8)
axs[0, 1].yaxis.set_tick_params(labelsize=8)
sns.swarmplot(y='local_density',
x='title',
data=BN_ldens,
size=1,
ax=axs[1, 1])
axs[1, 1].set_ylim([0, 6])
axs[1, 1].set_yticks([0, 2, 4, 6])
axs[1, 1].yaxis.label.set_size(8)
axs[1, 1].set_xlabel('')
axs[1, 1].yaxis.set_tick_params(labelsize=8)
axs[1, 1].xaxis.set_tick_params(labelsize=8)
plt.setp(axs[0, 0].get_xticklabels(), visible=False)
plt.setp(axs[0, 1].get_xticklabels(), visible=False)
fig.align_ylabels(axs[:, :])
plt.xticks(rotation=-20)
plt.subplots_adjust(hspace=0.12, wspace=0.1)
plt.savefig(os.path.join(output_dir, 'density.pdf'),
dpi=500,
bbox_inches='tight')
plt.show()
def main():
opentraj_root = sys.argv[1]
output_dir = sys.argv[2]
all_names = [
'ETH-Univ', 'ETH-Hotel', 'UCY-Zara', 'UCY-Univ', 'SDD-Coupa',
'SDD-bookstore', 'SDD-deathCircle', 'GC', 'InD-1', 'InD-2', 'KITTI',
'LCas-Minerva', 'WildTrack', 'Edinburgh', 'BN-1d-w180', 'BN-2d-w160'
]
datasets = get_datasets(opentraj_root, all_names)
#interaction range
upperbound = 3
thre = 3
lowerbound = 0.2
datasets_ttc = []
datasets_dca = []
datasets_collision_energy = []
# Get all datasets
for name in all_names:
dataset = datasets[name]
print("reading", name)
all_frames = dataset.get_frames()
#calculate and store ttc
all_trajs = dataset.get_trajectories("pedestrian")
trajlets = {}
#calculate and store ttc,dca,collision energy
avg_ttc, avg_dca = ttc(all_frames, name, trajlets)
allttc_pd = pd.DataFrame(data=np.zeros((len(avg_ttc), 2)),
columns=['name', 'ttc'])
allttc_pd.iloc[:, 0] = [name for i in range(len(avg_ttc))]
allttc_pd.iloc[:, 1] = avg_ttc
datasets_ttc.append(avg_ttc)
outputFile = output_dir + "/" + name + '_ttc.h5'
fw = open(outputFile, 'wb')
pickle.dump(allttc_pd, fw)
fw.close()
alldca_pd = pd.DataFrame(data=np.zeros((len(avg_dca), 2)),
columns=['name', 'dca'])
alldca_pd.iloc[:, 0] = [name for i in range(len(avg_dca))]
alldca_pd.iloc[:, 1] = avg_dca
datasets_dca.append(avg_dca)
outputFile = output_dir + "/" + name + '_dca.h5'
fw = open(outputFile, 'wb')
pickle.dump(alldca_pd, fw)
fw.close()
all_E = energy(avg_ttc, upperbound, lowerbound)
collision_energy = pd.DataFrame(data=np.zeros((len(all_E), 2)),
columns=['name', 'collision_energy'])
collision_energy.iloc[:, 0] = [name for i in range(len(all_E))]
collision_energy.iloc[:, 1] = all_E
datasets_collision_energy.append(all_E)
outputFile = output_dir + "/" + name + '_collisionEnergy.h5'
fw = open(outputFile, 'wb')
pickle.dump(collision_energy, fw)
fw.close()
# down-sample each group.
ttc_d = histogram_sampler(datasets_ttc, max_n_samples=500, n_bins=60)
dca_d = histogram_sampler(datasets_dca, max_n_samples=500, n_bins=60)
E_d = histogram_sampler(datasets_collision_energy,
max_n_samples=500,
n_bins=60)
# put samples in a DataFrame (required for seaborn plots)
df_ttc = pd.concat([
pd.DataFrame({
'title': all_names[ii],
'ttc': ttc_d[ii],
}) for ii in range(len(all_names))
])
df_dca = pd.concat([
pd.DataFrame({
'title': all_names[ii],
'dca': dca_d[ii],
}) for ii in range(len(all_names))
])
df_E = pd.concat([
pd.DataFrame({
'title': all_names[ii],
'collision_energy': E_d[ii],
}) for ii in range(len(all_names))
])
print("making plots ...")
sns.set(style="whitegrid")
fig, (ax1, ax2, ax3) = plt.subplots(3, 1, figsize=(12, 3), sharex=True)
sns.swarmplot(y='ttc', x='title', data=df_ttc, size=1, ax=ax1)
ax1.set_xlabel('')
ax1.set_yticks([0, 2.0, 4.0, 6.0])
ax1.xaxis.set_tick_params(labelsize=8)
ax1.yaxis.label.set_size(8)
ax1.yaxis.set_tick_params(labelsize=8)
sns.swarmplot(y='dca', x='title', data=df_dca, size=1, ax=ax2)
ax2.set_xlabel('')
ax2.xaxis.set_tick_params(labelsize=8)
ax2.yaxis.label.set_size(8)
ax2.yaxis.set_tick_params(labelsize=8)
sns.swarmplot(y='collision_energy', x='title', data=df_E, size=1, ax=ax3)
ax3.set_xlabel('')
plt.xticks(rotation=-20)
ax3.set_yticks([0, 4, 8, 12])
ax3.xaxis.set_tick_params(labelsize=8)
ax3.yaxis.label.set_size(8)
ax3.yaxis.set_tick_params(labelsize=8)
plt.subplots_adjust(hspace=0.1)
fig.align_ylabels()
plt.savefig(os.path.join(output_dir, 'collision.pdf'),
dpi=500,
bbox_inches='tight')
plt.show()