Spaces:
Sleeping
Sleeping
File size: 14,390 Bytes
e75a247 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 |
import matplotlib.pyplot as plt
import torch
import os
import numpy as np
import matplotlib.colors as mcolors
from matplotlib import cm
from sklearn.metrics import confusion_matrix
from src.plotting.histograms import score_histogram, confusion_matrix_plot
from src.plotting.plot_coordinates import plot_coordinates
from src.layers.object_cond import calc_eta_phi
def plot_event_comparison(event, ax=None, special_pfcands_size=1, special_pfcands_color="gray"):
eta_dq = event.matrix_element_gen_particles.eta
phi_dq = event.matrix_element_gen_particles.phi
pt_dq = event.matrix_element_gen_particles.pt
eta = event.pfcands.eta
phi = event.pfcands.phi
pt = event.pfcands.pt
mapping = event.pfcands.pf_cand_jet_idx.int().tolist()
print("N jets:", len(event.jets))
genjet_eta = event.genjets.eta
genjet_phi = event.genjets.phi
genjet_pt = event.genjets.pt
if ax is None:
fig, ax = plt.subplots(1, 2, figsize=(10, 5))
# plot eta, phi and the size of circles proportional to p_t. The colors should be either gray (if not in mapping) or some other color that 'represents' the identified jet
colorlist = ["red", "green", "blue", "purple", "orange", "yellow", "black", "pink", "cyan", "brown", "black", "black", "black", "gray"]
colors = []
for i in range(len(eta)):
colors.append(colorlist[mapping[i]])
colors = np.array(colors)
is_special = (event.pfcands.pid.abs() < 4)
#markers = ["." if not is_special[i] else "v" for i in range(len(eta))]
#ax[0].scatter(eta, phi, s=pt, c=colors)
ax[0].scatter(eta[is_special], phi[is_special], s=pt[is_special]*special_pfcands_size, c=special_pfcands_color, marker="v")
ax[0].scatter(eta[~is_special], phi[~is_special], s=pt[~is_special], c=colors[~is_special])
ax[0].scatter(eta_dq, phi_dq, s=pt_dq, c="red", marker="^", alpha=1.0)
ax[0].scatter(genjet_eta, genjet_phi, marker="*", s=genjet_pt, c="blue", alpha=1.0)
#eta_special = event.special_pfcands.eta
#phi_special = event.special_pfcands.phi
#pt_special = event.special_pfcands.pt
#print("N special PFCands:", len(eta_special))
#ax[0].scatter(eta_special, phi_special, s=pt_special*special_pfcands_size, c=special_pfcands_color, marker="v")
# "special" PFCands - electrons, muons, photons satisfying certain criteria
# Display the jets as a circle with R=0.5
jet_eta = event.jets.eta
jet_phi = event.jets.phi
for i in range(len(jet_eta)):
circle = plt.Circle((jet_eta[i], jet_phi[i]), 0.5, color="red", fill=False)
ax[0].add_artist(circle)
ax[0].set_xlabel(r"$\eta$")
ax[0].set_ylabel(r"$\phi$")
ax[0].set_title("PFCands with Jets")
if event.fatjets is not None:
colors = []
for i in range(len(eta)):
colors.append(colorlist[mapping[i]])
colors = np.array(colors)
is_special = (event.pfcands.pid.abs() < 4)
ax[1].scatter(eta[is_special], phi[is_special], s=pt[is_special] * special_pfcands_size,
c=colors[is_special], marker="v")
ax[1].scatter(eta[~is_special], phi[~is_special], s=pt[~is_special], c=colors[~is_special])
ax[1].scatter(eta_dq, phi_dq, s=pt_dq, c="red", marker="^", alpha=1.0)
ax[1].scatter(genjet_eta, genjet_phi, marker="*", s=genjet_pt, c="blue", alpha=1.0)
ax[1].set_xlabel(r"$\eta$")
ax[1].set_ylabel(r"$\phi$")
ax[1].set_title("PFCands with FatJets")
# Plot the fatjets as a circle with R=0.8 around the center of the fatjet
fatjet_eta = event.fatjets.eta
fatjet_phi = event.fatjets.phi
fatjet_R = 0.8
for i in range(len(fatjet_eta)):
circle = plt.Circle((fatjet_eta[i], fatjet_phi[i]), fatjet_R, color="red", fill=False)
ax[1].add_artist(circle)
# even aspect ratio
ax[1].set_aspect("equal")
ax[0].set_aspect("equal")
if ax is not None:
fig.tight_layout()
return fig
def plot_event(event, colors="gray", custom_coords=None, ax=None, jets=True, pfcands="pfcands"):
# plots event onto the specified ax.
# :colors: color of the pfcands
# :colors_special: color of the special pfcands
# :ax: matplotlib ax object to plot onto
# :custom_coords: Plot eta and phi from custom_coords instead of event.pfcands.
make_fig = ax is None
if ax is None:
fig, ax = plt.subplots(1, 1, figsize=(5, 5))
eta = getattr(event, pfcands).eta
phi = getattr(event, pfcands).phi
pt = getattr(event, pfcands).pt
#eta_special = event.special_pfcands.eta
#phi_special = event.special_pfcands.phi
#pt_special = event.special_pfcands.pt
if custom_coords:
eta = custom_coords[0]
phi = custom_coords[1]
#if len(eta_special):
# eta_special = eta[-len(eta_special):]
# phi_special = phi[-len(phi_special):]
# eta = eta[:-len(eta_special)]
# phi = phi[:-len(eta_special)]
#genjet_eta = event.genjets.eta
#genjet_phi = event.genjets.phi
#genjet_pt = event.genjets.pt
#if len(eta_special):
# colors_special = colors[-len(eta_special):]
# colors = colors[:-len(eta_special)]
# print("Colors_special", colors_special)
# assert len(colors) == len(phi)
# assert len(colors_special) == len(eta_special)
ax.scatter(eta, phi, s=pt, c=colors, alpha=0.7)
if hasattr(event, "matrix_element_gen_particles") and event.matrix_element_gen_particles is not None:
eta_dq = event.matrix_element_gen_particles.eta
phi_dq = event.matrix_element_gen_particles.phi
pt_dq = event.matrix_element_gen_particles.pt
ax.scatter(eta_dq, phi_dq, s=pt_dq, c="red", marker="^", alpha=0.5) # Dark quarks
#ax.scatter(genjet_eta, genjet_phi, marker="*", s=genjet_pt, c="blue", alpha=0.5)
#if len(eta_special):
# ax.scatter(eta_special, phi_special, s=pt_special, c=colors_special, marker="v")
if jets:
#jet_eta = event.fatjets.eta
#jet_phi = event.fatjets.phi
#for i in range(len(jet_eta)):
# circle = plt.Circle((jet_eta[i], jet_phi[i]), 0.8, color="red", fill=False)
# ax.add_artist(circle)
if hasattr(event, "model_jets") and event.model_jets is not None:
model_jet_eta = event.model_jets.eta
model_jet_phi = event.model_jets.phi
obj_score = None
if hasattr(event.model_jets, "obj_score"):
obj_score = event.model_jets.obj_score
for i in range(len(model_jet_eta)):
circle = plt.Circle((model_jet_eta[i], model_jet_phi[i]), 0.77, color="blue", fill=False, alpha=.7)
ax.add_artist(circle)
# plot text with obj score
if obj_score is not None:
ax.text(model_jet_eta[i]+0.2, model_jet_phi[i]-0.2, "o.s.=" + str(round(torch.sigmoid(obj_score[i]).item(), 2)), color="gray", fontsize=10, alpha=0.5)
if hasattr(event, "fastjet_jets") and event.fastjet_jets is not None:
fj_r = 0.8
model_jet_eta = event.fastjet_jets[fj_r].eta
model_jet_phi = event.fastjet_jets[fj_r].phi
for i in range(len(model_jet_eta)):
circle = plt.Circle((model_jet_eta[i], model_jet_phi[i]), 0.74, color="green", fill=False, alpha=.7)
ax.add_artist(circle)
ax.set_xlabel(r"$\eta$")
ax.set_ylabel(r"$\phi$")
ax.set_aspect("equal")
if make_fig:
fig.tight_layout()
return fig
def get_idx_for_event(obj, i):
return obj.batch_number[i], obj.batch_number[i+1]
def get_labels_jets(b, pfcands, jets):
# b: Batch of events
R = 0.8
labels = torch.zeros(len(pfcands)).long()
for i in range(len(b)):
s, e = get_idx_for_event(jets, i)
dq_eta = jets.eta[s:e]
dq_phi = jets.phi[s:e]
if s == e:
continue
s, e = get_idx_for_event(pfcands, i)
pfcands_eta = pfcands.eta[s:e]
pfcands_phi = pfcands.phi[s:e]
# calculate the distance matrix between each dark quark and pfcands
dist_matrix = torch.cdist(
torch.stack([dq_eta, dq_phi], dim=1),
torch.stack([pfcands_eta, pfcands_phi], dim=1),
p=2
)
dist_matrix = dist_matrix.T
closest_quark_dist, closest_quark_idx = dist_matrix.min(dim=1)
closest_quark_idx[closest_quark_dist > R] = -1
labels[s:e] = closest_quark_idx
return (labels >= 0).float()
def plot_batch_eval_OC(event_batch, y_true, y_pred, batch_idx, filename, args, batch, dropped_batches):
# Plot the batch, together with nice colors with object condensation GT and betas
max_events = 5
sz = 10
assert len(y_true) == len(y_pred), f"y_true: {len(y_true)}, y_pred: {len(y_pred)}"
if args.beta_type == "pt+bc":
n_columns = 6
y_true_bc = (y_true >= 0).int()
#score_histogram(y_true_bc, y_pred[:, 3]).savefig(os.path.join(os.path.dirname(filename), "binary_classifier_scores.pdf"))
#score_histogram(y_true_bc, (event_batch.pfcands.pf_cand_jet_idx >= 0).float()).savefig(
# os.path.join(os.path.dirname(filename), "binary_classifier_scores_AK8.pdf"))
#score_histogram(y_true_bc, get_labels_jets(event_batch, event_batch.pfcands, event_batch.fatjets)).savefig(
# os.path.join(os.path.dirname(filename), "binary_classifier_scores_radius_FatJets.pdf"))
#score_histogram(y_true_bc, get_labels_jets(event_batch, event_batch.pfcands, event_batch.genjets)).savefig(
# os.path.join(os.path.dirname(filename), "binary_classifier_scores_radius_GenJets.pdf"))
#fig, ax = plt.subplots(1, 3, figsize=(3*sz/2, sz/2))
#confusion_matrix_plot(y_true_bc, y_pred[:, 3] > 0.5, ax[0])
#ax[0].set_title("Classifier (cut at 0.5)")
#confusion_matrix_plot(y_true_bc, get_labels_jets(event_batch, event_batch.pfcands, event_batch.fatjets), ax[2])
#ax[2].set_title("FatJets")
#confusion_matrix_plot(y_true_bc, get_labels_jets(event_batch, event_batch.pfcands, event_batch.genjets), ax[1])
#ax[1].set_title("GenJets")
#fig.tight_layout()
#fig.savefig(os.path.join(os.path.dirname(filename), "conf_matrices.pdf"))
else:
n_columns = 4
fig, ax = plt.subplots(max_events, n_columns, figsize=(n_columns * sz, sz * max_events))
# columns: Input coords, colored by beta ; Input coords, colored by GT labels; model coords, colored by beta; model coords, colored by GT labels
print("N events")
for i in range(event_batch.n_events):
if i >= max_events:
break
if i not in dropped_batches:
continue
event = event_batch[i]
filt = batch_idx == i
y_true_event = y_true[filt]
y_pred_event = y_pred[filt]
if args.beta_type == "default":
betas = y_pred_event[filt, 3]
elif args.beta_type == "pt":
betas = event.pfcands.pt
elif args.beta_type == "pt+bc":
betas = event.pfcands.pt
classifier_labels = y_pred_event[:, 3]
p_xyz = y_pred_event[:, :3]
if y_pred_event.shape[1] == 5:
p_xyz = y_pred_event[:, 1:4]
e = y_pred_event[:, 0]
#lorentz_invariant = e**2 - p_xyz.norm(dim=1)**2
#lorentz_invariant_inputs = event.pfcands.E ** 2 - event.pfcands.pxyz.norm(dim=1) ** 2
plot_coordinates(event.pfcands.pxyz, pt=event.pfcands.pt, tidx=y_true_event,
outdir=os.path.dirname(filename),
filename="input_coords_batch_" + str(batch) + "_event_" + str(i) + ".html")
plot_coordinates(p_xyz, pt=event.pfcands.pt, tidx=y_true_event,
outdir=os.path.dirname(filename),
filename="model_coords_batch_" + str(batch) + "_event_" + str(i) + ".html")
y_true_event = y_true_event.tolist()
clist = ['#1f78b4', '#b3df8a', '#33a02c', '#fb9a99', '#e31a1c', '#fdbe6f', '#ff7f00', '#cab2d6', '#6a3d9a', '#ffff99', '#b15928']
colors = {
-1: "gray",
0: clist[0],
1: clist[1],
2: clist[2],
3: clist[3]
}
eta, phi = calc_eta_phi(p_xyz, return_stacked=False)
plot_event(event, colors=plt.cm.brg(betas), ax=ax[i, 0])
cbar = plt.colorbar(mappable=cm.ScalarMappable(cmap=plt.cm.brg), ax=ax[i, 0]) # How to specify the palette?
ax[i, 0].set_title(r"input coords, $\beta$ colors")
cbar.set_label(r"$\beta$")
plot_event(event, colors=[colors[i] for i in y_true_event], ax=ax[i, 1])
ax[i, 1].set_title("input coords, GT colors")
plot_event(event, custom_coords=[eta, phi], colors=plt.cm.brg(betas), ax=ax[i, 2], jets=False)
#assert betas.min() >= 0 and betas.max() <= 1
ax[i, 2].set_title(r"model coords, $\beta$ colors")
cbar = plt.colorbar(mappable=cm.ScalarMappable(cmap=plt.cm.brg), ax=ax[i, 2])
ax[i, 3].set_title("model coords, GT colors")
cbar.set_label(r"$\beta$")
plot_event(event, custom_coords=[eta, phi], colors=[colors[i] for i in y_true_event], ax=ax[i, 3], jets=False)
if args.beta_type == "pt+bc":
# Create a custom colormap from light gray to dark green
colors = [(0.9, 0.9, 0.9), (0.0, 0.5, 0.0)] # RGB for light gray and dark green
cmap_name = "lightgray_to_darkgreen"
custom_cmap = mcolors.LinearSegmentedColormap.from_list(cmap_name, colors)
plot_event(event, custom_coords=[eta, phi], colors=custom_cmap(classifier_labels), ax=ax[i, 5], jets=False)
ax[i, 5].set_title(r"model coords, BC label colors")
cbar = plt.colorbar(mappable=cm.ScalarMappable(cmap=custom_cmap), ax=ax[i, 5])
cbar.set_label("Classifier score")
plot_event(event, colors=custom_cmap(classifier_labels), ax=ax[i, 4], jets=False)
ax[i, 4].set_title(r"input coords, BC label colors")
cbar = plt.colorbar(mappable=cm.ScalarMappable(cmap=custom_cmap), ax=ax[i, 4])
cbar.set_label("Classifier score")
print("Saving eval figure to", filename)
fig.tight_layout()
fig.savefig(filename)
fig.clear()
plt.close(fig)
|