Skip to content

Commit bb5b1b8

Browse files
rhoadesScholarrhoadesScholar
committed
boxplots
1 parent 5493d18 commit bb5b1b8

File tree

2 files changed

+58
-54
lines changed

2 files changed

+58
-54
lines changed

‎experiments/ieee-isbi-2023/03_evaluate/inspect_tests.py‎

Lines changed: 44 additions & 41 deletions
Original file line numberDiff line numberDiff line change
@@ -1,25 +1,34 @@
11
# %%
22
from raygun.evaluation.inspect_tests import *
33

4+
45
# switch to svg backend
5-
# matplotlib.use("svg")
6+
matplotlib.use("svg")
67
# update latex preamble
7-
# plt.rcParams.update(
8-
# {
9-
# "svg.fonttype": "path",
10-
# # "font.family": "sans-serif",
11-
# # "font.sans-serif": "AvenirNextLTPro", # ["Avenir", "AvenirNextLTPro", "Avenir Next LT Pro", "AvenirNextLTPro-Regular", 'UniversLTStd-Light', 'Verdana', 'Helvetica']
12-
# "path.simplify": True,
13-
# # "text.usetex": True,
14-
# # "pgf.rcfonts": False,
15-
# # "pgf.texsystem": 'pdflatex', # default is xetex
16-
# # "pgf.preamble": [
17-
# # r"\usepackage[T1]{fontenc}",
18-
# # r"\usepackage{mathpazo}"
19-
# # ]
20-
# }
21-
# )
22-
8+
plt.rcParams.update(
9+
{
10+
"svg.fonttype": "path",
11+
# "font.family": "sans-serif",
12+
# "font.sans-serif": [
13+
# "Avenir",
14+
# "AvenirNextLTPro",
15+
# "Avenir Next LT Pro",
16+
# "AvenirNextLTPro-Regular",
17+
# "UniversLTStd-Light",
18+
# "Verdana",
19+
# "Helvetica",
20+
# ],
21+
"path.simplify": True,
22+
# "text.usetex": True,
23+
# "pgf.rcfonts": False,
24+
# "pgf.texsystem": 'pdflatex', # default is xetex
25+
# "pgf.preamble": [
26+
# r"\usepackage[T1]{fontenc}",
27+
# r"\usepackage{mathpazo}"
28+
# ]
29+
"font.size": 14,
30+
}
31+
)
2332
# %%
2433
paths = [
2534
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2023/03_evaluate/test_eval1_metrics.json",
@@ -122,19 +131,6 @@
122131
import pandas as pd
123132
import seaborn as sns
124133

125-
# trains = [
126-
# "link",
127-
# "split",
128-
# "real30nm",
129-
# # "real90nm",
130-
# ] # set([keys[0] for keys in list(sums.values())[0].keys()])
131-
# predicts = [
132-
# # "real30nm",
133-
# "link",
134-
# "split",
135-
# "real90nm",
136-
# ] # set([keys[1] for keys in list(sums.values())[0].keys()])
137-
138134

139135
def get_df(sums, pairs, metric="voi"):
140136
df = pd.DataFrame()
@@ -175,6 +171,7 @@ def box_plots(
175171

176172
df = get_df(sums, pairs.values(), metric)
177173
sns.boxplot(
174+
# sns.violinplot(
178175
ax=axes[c],
179176
data=df,
180177
x=df.index,
@@ -196,8 +193,9 @@ def box_plots(
196193
axes[c].plot(
197194
range(-1, x + 1), [means[metric][baseline]] * (x + 2), style, label=name
198195
)
196+
axes[c].text(x - 0.45, means[metric][baseline], name, color=style[0])
199197
axes[c].set_xlim(-0.5, x - 0.5)
200-
axes[c].legend()
198+
# axes[c].legend()
201199
fig.tight_layout()
202200

203201
return fig
@@ -212,13 +210,14 @@ def box_plots(
212210
}
213211

214212
baselines = {
215-
"Naive": (("real30nm", "real90nm"), "r--"),
213+
"Naïve": (("real30nm", "real90nm"), "r--"),
216214
"Paired": (("real90nm", "real90nm"), "g--"),
217215
}
218216

219217
fig = box_plots(pairs, baselines)
220-
fig.savefig("boxplots_compare_all.png", dpi=300)
221-
218+
# fig.savefig("boxplots_compare_all.png", dpi=300)
219+
# fig.savefig("boxplots_compare_all.svg", dpi=300)
220+
fig
222221
# %%
223222

224223
pairs = {
@@ -229,7 +228,7 @@ def box_plots(
229228
}
230229

231230
baselines = {
232-
"Naive": (("real30nm", "real90nm"), "r--"),
231+
"Naïve": (("real30nm", "real90nm"), "r--"),
233232
"Paired": (("real90nm", "real90nm"), "g--"),
234233
}
235234

@@ -243,11 +242,14 @@ def box_plots(
243242
}
244243

245244
baselines = {
246-
"Naive": (("real30nm", "real90nm"), "r--"),
245+
"Naïve": (("real30nm", "real90nm"), "r--"),
247246
}
248247

249248
fig = box_plots(pairs, baselines)
250-
fig.savefig("boxplots_compare_split.png", dpi=300)
249+
# fig.savefig("boxplots_compare_split.png", dpi=300)
250+
fig.savefig("boxplots_compare_split.svg", dpi=300)
251+
fig
252+
251253
# %%
252254

253255
pairs = {
@@ -256,7 +258,7 @@ def box_plots(
256258
}
257259

258260
baselines = {
259-
"Naive": (("real30nm", "real90nm"), "r--"),
261+
"Naïve": (("real30nm", "real90nm"), "r--"),
260262
"Paired": (("real90nm", "real90nm"), "g--"),
261263
}
262264

@@ -270,10 +272,11 @@ def box_plots(
270272
}
271273

272274
baselines = {
273-
"Naive": (("real30nm", "real90nm"), "r--"),
275+
"Naïve": (("real30nm", "real90nm"), "r--"),
274276
}
275277

276278
fig = box_plots(pairs, baselines)
277-
fig.savefig("boxplots_compare_link.png", dpi=300)
278-
279+
# fig.savefig("boxplots_compare_link.png", dpi=300)
280+
fig.savefig("boxplots_compare_link.svg", dpi=300)
281+
fig
279282
# %%

‎experiments/ieee-isbi-2023/03_evaluate/qualitative_figs.py‎

Lines changed: 14 additions & 13 deletions
Original file line numberDiff line numberDiff line change
@@ -30,6 +30,7 @@
3030
# r"\usepackage[T1]{fontenc}",
3131
# r"\usepackage{mathpazo}"
3232
# ]
33+
# "font.size": 20,
3334
}
3435
)
3536

@@ -134,10 +135,10 @@ def show_images(
134135
"Real 30nm": "volumes/raw_30nm",
135136
"Real 90nm": "volumes/interpolated_90nm_aligned",
136137
},
137-
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2022/01_cycleGAN/link/seed42/training_0.n5": {
138+
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2023/01_cycleGAN/link/seed42/training_0.n5": {
138139
"Link: Fake 90nm (best)": "volumes/raw_30nm_netG2_62000" # picked based on final test performance
139140
},
140-
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2022/01_cycleGAN/split/seed42/training_0.n5": {
141+
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2023/01_cycleGAN/split/seed42/training_0.n5": {
141142
"Split: Fake 90nm (best)": "volumes/raw_30nm_netG2_36000" # picked based on final test performance
142143
},
143144
}
@@ -167,10 +168,10 @@ def show_images(
167168
"Real 30nm": "volumes/raw_30nm",
168169
"Real 90nm": "volumes/interpolated_90nm_aligned",
169170
},
170-
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2022/01_cycleGAN/link/seed13/eval_1.n5": {
171+
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2023/01_cycleGAN/link/seed13/eval_1.n5": {
171172
"Link: Fake 30nm (best)": "volumes/interpolated_90nm_aligned_netG1_46000" # picked based on final test performance
172173
},
173-
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2022/01_cycleGAN/split/seed42/eval_1.n5": {
174+
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2023/01_cycleGAN/split/seed42/eval_1.n5": {
174175
"Split: Fake 30nm (best)": "volumes/interpolated_90nm_aligned_netG1_36000" # picked based on final test performance
175176
},
176177
}
@@ -188,27 +189,27 @@ def show_images(
188189
"/nrs/funke/rhoadesj/data/XNH/CBv/GT/CBvTopGT/eval_1.n5": {
189190
"Real 30nm": "volumes/raw_30nm"
190191
},
191-
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2022/03_evaluate/train_real/30nm/predict_real/30nm/eval_1.n5": {
192+
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2023/03_evaluate/train_real/30nm/predict_real/30nm/eval_1.n5": {
192193
"Predicted\nLocal Shape\nDescriptors": "pred_lsds",
193194
"Predicted\nAffinities": "pred_affs",
194195
"Predicted\nSegmentation": "segment",
195196
},
196197
},
197198
{ # Train on Real 30nm, Predict on Link-Fake 30nm
198-
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2022/01_cycleGAN/link/seed13/eval_1.n5": {
199+
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2023/01_cycleGAN/link/seed13/eval_1.n5": {
199200
"Link: Fake 30nm (best)": "volumes/interpolated_90nm_aligned_netG1_46000"
200201
},
201-
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2022/03_evaluate/train_real/30nm/predict_link/seed13/eval_1.n5": {
202+
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2023/03_evaluate/train_real/30nm/predict_link/seed13/eval_1.n5": {
202203
"Predicted\nLocal Shape\nDescriptors": "pred_lsds",
203204
"Predicted\nAffinities": "pred_affs",
204205
"Predicted\nSegmentation": "segment",
205206
},
206207
},
207208
{ # Train on Real 30nm, Predict on Split-Fake 30nm
208-
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2022/01_cycleGAN/split/seed42/eval_1.n5": {
209+
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2023/01_cycleGAN/split/seed42/eval_1.n5": {
209210
"Split: Fake 30nm (best)": "volumes/interpolated_90nm_aligned_netG1_36000"
210211
},
211-
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2022/03_evaluate/train_real/30nm/predict_split/seed42/eval_1.n5": {
212+
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2023/03_evaluate/train_real/30nm/predict_split/seed42/eval_1.n5": {
212213
"Predicted\nLocal Shape\nDescriptors": "pred_lsds",
213214
"Predicted\nAffinities": "pred_affs",
214215
"Predicted\nSegmentation": "segment",
@@ -218,7 +219,7 @@ def show_images(
218219
"/nrs/funke/rhoadesj/data/XNH/CBv/GT/CBvTopGT/eval_1.n5": {
219220
"Real 90nm\n(Trained on:\nreal 30nm)": "volumes/interpolated_90nm_aligned"
220221
},
221-
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2022/03_evaluate/train_real/30nm/predict_real/90nm/eval_1.n5": {
222+
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2023/03_evaluate/train_real/30nm/predict_real/90nm/eval_1.n5": {
222223
"Predicted\nLocal Shape\nDescriptors": "pred_lsds",
223224
"Predicted\nAffinities": "pred_affs",
224225
"Predicted\nSegmentation": "segment",
@@ -228,7 +229,7 @@ def show_images(
228229
"/nrs/funke/rhoadesj/data/XNH/CBv/GT/CBvTopGT/eval_1.n5": {
229230
"Real 90nm\n(Trained on:\nreal 90nm)": "volumes/interpolated_90nm_aligned"
230231
},
231-
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2022/03_evaluate/train_real/90nm/predict_real90nm/eval_1.n5": {
232+
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2023/03_evaluate/train_real/90nm/predict_real90nm/eval_1.n5": {
232233
"Predicted\nLocal Shape\nDescriptors": "pred_lsds",
233234
"Predicted\nAffinities": "pred_affs",
234235
"Predicted\nSegmentation": "segment",
@@ -238,7 +239,7 @@ def show_images(
238239
"/nrs/funke/rhoadesj/data/XNH/CBv/GT/CBvTopGT/eval_1.n5": {
239240
"Real 90nm\n(Trained on:\nLink-Fake 90nm)": "volumes/interpolated_90nm_aligned"
240241
},
241-
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2022/03_evaluate/train_link/seed42/predict_real90nm/eval_1.n5": {
242+
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2023/03_evaluate/train_link/seed42/predict_real90nm/eval_1.n5": {
242243
"Predicted\nLocal Shape\nDescriptors": "pred_lsds",
243244
"Predicted\nAffinities": "pred_affs",
244245
"Predicted\nSegmentation": "segment",
@@ -248,7 +249,7 @@ def show_images(
248249
"/nrs/funke/rhoadesj/data/XNH/CBv/GT/CBvTopGT/eval_1.n5": {
249250
"Real 90nm\n(Trained on:\nSplit-Fake 90nm)": "volumes/interpolated_90nm_aligned"
250251
},
251-
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2022/03_evaluate/train_split/seed42/predict_real90nm/eval_1.n5": {
252+
"/nrs/funke/rhoadesj/raygun/experiments/ieee-isbi-2023/03_evaluate/train_split/seed42/predict_real90nm/eval_1.n5": {
252253
"Predicted\nLocal Shape\nDescriptors": "pred_lsds",
253254
"Predicted\nAffinities": "pred_affs",
254255
"Predicted\nSegmentation": "segment",

0 commit comments

Comments
 (0)