-
Notifications
You must be signed in to change notification settings - Fork 11
/
plotting.py
336 lines (320 loc) · 13.5 KB
/
plotting.py
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
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
import json
import os
import statistics
from collections import Counter
import plotly
import plotly.figure_factory as ff
import plotly.graph_objects as go
import torch
plotly.io.orca.config.executable = "PATH/TO/orca"
'''
Notes: 73% of videos pass the filter
3.2 seconds and up get annotated
>40 seconds ignored
failure location not in [1,99]% ignored
'''
class MaybeFigure:
def __init__(self, name, fig_names, figs, gen_fn):
self.ok = fig_names is None or name in fig_names
self.name = name
self.figs = figs
self.gen_fn = gen_fn
def __enter__(self):
return self
def assign(self, d):
if self.ok:
self.figs[self.name] = d
def gen(self):
self.assign(self.gen_fn(self.name))
def __exit__(self, type, value, traceback):
if self.ok:
print(f'processed {self.name}')
class FigureBuilder:
def __init__(self, anns, confusion=None, filetype='pdf', basepath="PATH/TO/tmp",
ok_names=None):
self.anns = anns
self.filetype = filetype
self.confusion = confusion
self.basepath = basepath
self.ok_names = ok_names
self.all_names = ['lbl_rel_dist', 'lbl_rel_stdev', 'vid_len', 'kinetics_dist', 'places_dist', 'category_err']
self.figs = {}
self.default_layout_kwargs = dict(margin=dict(l=10, r=10, t=55, b=10), height=350,
font=dict(size=26, family='Inter'))
self.default_layout_x_kwargs = dict(tickfont=dict(size=20))
self.default_layout_y_kwargs = dict(tickfont=dict(size=20))
self.default_title_kwargs = dict(x=0.5, xref='container')
self.maybe_figure = lambda name: MaybeFigure(name, self.ok_names, self.figs, gen_fn=self.gen_fig)
def __call__(self, *args, **kwargs):
self.gen_figs()
def gen_fig(self, name):
if name == 'confusion':
M = self.confusion
M /= M.sum(dim=1).unsqueeze(1)
M *= 100
M = M.flip([0]).tolist()
for i in range(len(M)):
for j in range(len(M[i])):
M[i][j] = round(M[i][j], 2)
fig = ff.create_annotated_heatmap(z=M, x=['Before failure', 'At failure', 'After failure'],
y=['Before failure', 'At failure', 'After failure'], colorscale='Viridis')
return {
'fig': fig,
'title': 'Classification confusion matrix',
'xtitle': 'Ground truth',
'ytitle': '<span>Prediction<br> </span>'
}
elif name == 'lbl_rel_dist':
rel_dist_data = {k: statistics.median(v['rel_t']) for k, v in self.anns.items() if
0.01 <= statistics.median(v['rel_t']) <= 0.99}
fig = ff.create_distplot([[_ for _ in rel_dist_data.values()]], [''], show_rug=False, bin_size=0.02,
curve_type='normal', histnorm='probability', show_curve=False)
fig.update_traces(marker_line_color='rgb(255, 127, 14)', marker_color='rgb(255, 127, 14)',
marker_line_width=1,
opacity=0.6)
return {
'fig': fig,
'title': '(b) Failure label times',
'xtitle': 'Time (% of duration)',
'ytitle': 'Frequency',
'layout_kwargs': {
'showlegend': False
},
'layout_x_kwargs': {
'tickformat': '%'
},
'layout_y_kwargs': {
'tickformat': '.1%',
'rangemode': 'nonnegative'
# 'range': [0, 1]
},
'title_kwargs': {
}
}
elif name == 'lbl_rel_stdev':
rel_stdev_data = {k: statistics.stdev([max(0, _) for _ in v['rel_t']]) for k, v in self.anns.items() if
0.01 <= statistics.median(v['rel_t']) <= 0.99}
fig = ff.create_distplot([[_ for _ in rel_stdev_data.values()]], [''], show_rug=False, bin_size=0.01,
histnorm='probability', show_curve=False)
fig.update_traces(marker_line_color='rgb(44, 160, 44)', marker_color='rgb(44, 160, 44)',
marker_line_width=1,
opacity=0.6)
return {
'fig': fig,
'title': '(c) Failure label standard deviations',
'xtitle': 'Deviation (% of duration)',
'ytitle': 'Frequency',
'layout_kwargs': {
'showlegend': False
},
'layout_x_kwargs': {
'tickformat': '%'
},
'layout_y_kwargs': {
'tickformat': '%',
'rangemode': 'nonnegative'
# 'range': [0, 1]
},
'title_kwargs': {
}
}
elif name == 'lbl_stdev':
stdev_data = {k: statistics.stdev([max(0, _) for _ in v['t']]) for k, v in self.anns.items() if
0.01 <= statistics.median(v['rel_t']) <= 0.99}
fig = ff.create_distplot([[(_ if _ < 10 else 10) for _ in stdev_data.values()]], [''], show_rug=False,
bin_size=0.1,
histnorm='probability', show_curve=False)
fig.update_traces(marker_line_color='rgb(44, 160, 44)', marker_color='rgb(44, 160, 44)',
marker_line_width=1,
opacity=0.6)
return {
'fig': fig,
'title': 'Failure label standard deviations',
'xtitle': 'Standard deviation (seconds)',
'ytitle': 'Frequency',
'layout_kwargs': {
'showlegend': False
},
'layout_x_kwargs': {
# 'tickformat': '%'
},
'layout_y_kwargs': {
'tickformat': '.1%',
'rangemode': 'nonnegative'
# 'range': [0, 1]
},
'title_kwargs': {
}
}
elif name == 'vid_len':
vid_basepath = "PATH/TO/"
with open(os.path.join(vid_basepath, 'validcliplens.json'), 'r') as fff:
validcliplens = json.load(fff)
# ann_lens = {k: v['len'] for k, v in anns.items() if 0.01 <= statistics.median(v['rel_t']) <= 0.99}
return {
'fig': ff.create_distplot([[_ for _ in validcliplens if 3.2 <= _ < 30]], [''], show_rug=False,
bin_size=1,
histnorm='probability', show_curve=False),
'title': '(a) Video lengths',
'xtitle': 'Length (seconds)',
'ytitle': 'Frequency',
'layout_kwargs': {
'showlegend': False
},
'layout_y_kwargs': {
'tickformat': '%',
'rangemode': 'nonnegative'
# 'range': [0, 1]
},
'layout_x_kwargs': {
# 'range': [2, 40]
},
'title_kwargs': {
}
}
elif name == 'kinetics_dist':
kinetics_dist = torch.load("PATH/TO/kinetics_dist.pt")
kinetics_cls = torch.load("PATH/TO/kinetics_classes.pt")
kinetics_xs, kinetics_ys = zip(
*Counter({kinetics_cls[k][:25]: v for k, v in kinetics_dist.items()}).most_common())
fig = go.Figure([go.Bar(x=kinetics_xs, y=kinetics_ys)])
fig.update_traces(marker_line_color='rgb(214, 39, 40)', marker_color='rgb(214, 39, 40)',
marker_line_width=1,
opacity=0.6)
return {
'fig': fig,
'title': '(d) Video action classes',
'xtitle': None,
'ytitle': 'Frequency',
'layout_y_kwargs': {
'tickformat': '.1%',
# 'rangemode': 'nonnegative'
# 'range': [0, 1]
},
'layout_x_kwargs': {
'tickfont': dict(size=11)
# 'range': [2, 40]
},
'layout_kwargs': {
'xaxis_tickangle': 270,
'bargap': 0,
'height': 370,
'font': dict(size=18, family='Inter')
}
}
elif name == 'places_dist':
with open("PATH/TO/places_dist.json") as f:
places_dist = json.load(f)
with open("PATH/TO/categories_places365.txt") as f:
places_cls = f.readlines()
places_xs, places_ys = zip(
*Counter({' - '.join(places_cls[int(k)].split(' ')[0][3:].replace('_', ' ').split('/'))[:25]: v / sum(
places_dist.values()) for k, v in places_dist.items()}).most_common()[1:])
fig = go.Figure([go.Bar(x=places_xs, y=places_ys)])
fig.update_traces(marker_line_color='rgb(148, 103, 189)', marker_color='rgb(148, 103, 189)',
marker_line_width=1,
opacity=0.6)
return {
'fig': fig,
'title': '(e) Video scene classes',
'xtitle': None,
'ytitle': 'Frequency',
'layout_y_kwargs': {
'tickformat': '.1%',
# 'rangemode': 'nonnegative'
# 'range': [0, 1]
},
'layout_x_kwargs': {
'tickfont': dict(size=11)
# 'range': [2, 40]
},
'layout_kwargs': {
'xaxis_tickangle': 270,
'bargap': 0,
'height': 365,
'font': dict(size=18, family='Inter')
}
}
elif name == 'category_err':
categories = ["Multi-agent", "Single-agent", "Execution error", "Planning error", "Limited visibility",
"Unexpected", "Environmental", "Limited knowledge", "Limited skill"]
fig = go.Figure()
fig.add_trace(
go.Bar(name='Human', y=categories, x=[3.41, 4.58, 4.81, 14.4, 5.27, 3.26, 2.79, 6.15, 3.48],
orientation='h'))
fig.add_trace(
go.Bar(name='Kinetics fine-tuning', y=categories, x=[11, 11.9, 7.8, 5.8, 10.6, 14.8, 11.9, 10.5, 9.5],
orientation='h'))
fig.add_trace(
go.Bar(name='FPS fine-tuning', y=categories, x=[6.8, 13.4, 11.1, 12.4, 16.1, 21.3, 15.4, 13.1, 8.8],
orientation='h'))
# Change the bar mode
fig.update_layout(barmode='group', legend_orientation="h", legend=dict(y=-0.2), annotations=[
go.layout.Annotation(
x=0.5,
y=-0.15,
showarrow=False,
text="Error (%)",
xref="paper",
yref="paper",
font=dict(size=25, family='Inter')
)
])
# fig.show()
return {
'fig': fig,
'title': '',
'xtitle': '',
'ytitle': '',
'layout_y_kwargs': {
# 'tickformat': '.1%',
# 'rangemode': 'nonnegative'
# 'range': [0, 1]
'tickfont': dict(size=26)
},
'layout_x_kwargs': {
# 'tickfont': dict(size=11)
# 'range': [2, 40]
},
'layout_kwargs': {
# 'xaxis_tickangle': 90,
# 'bargap': 0,
'height': 650,
'margin': dict(l=0, r=10, t=10, b=200),
'font': dict(size=25, family='Inter')
}
}
def gen_figs(self):
for name in self.all_names:
with self.maybe_figure(name) as f:
f.gen()
for k, v in self.figs.items():
fig = v['fig']
fig.update_layout(
title=go.layout.Title(
text=v['title'],
**{**self.default_title_kwargs,
**v.get('title_kwargs', {})},
) if 'title' in v else None,
xaxis=go.layout.XAxis(
title=go.layout.xaxis.Title(
text=v['xtitle']
),
**{**self.default_layout_x_kwargs,
**v.get('layout_x_kwargs', {})},
),
yaxis=go.layout.YAxis(
title=go.layout.yaxis.Title(
text=v['ytitle']
),
**{**self.default_layout_y_kwargs,
**v.get('layout_y_kwargs', {})},
),
**{**self.default_layout_kwargs,
**v.get('layout_kwargs', {})},
)
fig.update_yaxes(automargin=True)
fig_fn = os.path.join(self.basepath, f'{k}.{self.filetype}')
print(f'writing {fig_fn}')
fig.write_image(fig_fn)
print('done!')