Repository navigation
Expand file tree
/
Copy pathplot.py
More file actions
371 lines (297 loc) · 14.9 KB
/
Copy pathplot.py
File metadata and controls
371 lines (297 loc) · 14.9 KB
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
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
"""
import os
import pandas as pd
# PATH = "corrected/dalle_eval/llava/vivid/data/"
# # combine all csv files in the folder
# combined_csv = pd.concat([pd.read_csv(f"{PATH}{f}") for f in os.listdir(PATH) if f.endswith(".csv")], ignore_index=True)
# df1 = pd.read_csv("corrected/dalle_street_vivid_llava.csv")
# df2 = pd.read_csv("corrected/dalle_street_natural_llava.csv")
# combined_csv = pd.concat([df1, df2], ignore_index=True)
# export to csv
# combined_csv.to_csv(f"corrected/dalle_street_llava.csv", index=False)
"""
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from matplotlib.patches import Rectangle
from sklearn.metrics import confusion_matrix, ConfusionMatrixDisplay
import json
import os
from tqdm import tqdm
import itertools
sns.set(font_scale=1.4)
sns.set_theme(style="whitegrid")
sns.set_context("paper", rc={"font.size":20,"axes.titlesize":20,"axes.labelsize":20})
DATA = "dollar" # "dalle"
MODEL = "llava" # "llava"
PATH = f"corrected/{DATA}_street_{MODEL}.csv"
#income plot
# Load JSON mapping file
with open('unsd_geoscheme.json') as f:
mappings = json.load(f)
# Flatten the JSON mapping for easier lookup
country_to_subregion = {}
subregion_to_continent = {}
for continent, regions in mappings.items():
for subregion, data in regions.items():
subregion_to_continent[subregion] = continent
for country in data['countries_included']:
country_to_subregion[country] = (subregion, continent)
# Load the CSV file
df = pd.read_csv(PATH)
# create a blank column for the predicted continent
df["predicted_continent"] = ""
if MODEL == "llava":
invalids = ["Africa", "Americas", "Asia", "Europe", "Oceania", "Antarctica", "Middle East", "ResponsibleAIPolicyViolation"]
else:
invalids = ["Africa", "Americas", "Asia", "Europe", "Oceania", "ResponsibleAIPolicyViolation"]
df.loc[df["predicted_continent"] == "", "predicted_continent"] = df.loc[df["predicted_continent"] == "", "response"].map(subregion_to_continent)
for invalid in invalids:
df.loc[df["response"].str.contains(invalid), "predicted_continent"] = invalid
print(df.head())
continents_to_plot = ["Africa", "Americas", "Asia", "Europe", "Oceania"]
# Calculate if the prediction is correct
df['correct_prediction'] = df['predicted_continent'] == df['true_continent']
# Filter the dataframe to keep only the relevant continents
continents_to_plot = ["Africa", "Americas", "Asia", "Europe", "Oceania"]
df = df[df['true_continent'].isin(continents_to_plot)]
# Calculate income quartiles for each continent separately
def calculate_income_bucket(group):
try:
return pd.qcut(group, q=4, labels=['Q1', 'Q2', 'Q3', 'Q4'], duplicates='drop')
except ValueError:
return pd.cut(group.rank(method='first'), bins=4, labels=['Q1', 'Q2', 'Q3', 'Q4'])
df['income_bucket'] = df.groupby('true_continent', group_keys=False)['income'].apply(calculate_income_bucket)
# Calculate the total and correct counts by continent and income bucket
total_counts = df.groupby(['true_continent', 'income_bucket']).size().reset_index(name='total_counts')
correct_counts = df[df['correct_prediction']].groupby(['true_continent', 'income_bucket']).size().reset_index(name='correct_counts')
# Merge the counts
merged_df = pd.merge(correct_counts, total_counts, on=['true_continent', 'income_bucket'], how='left')
# Calculate the normalized correct classification percentages
merged_df['normalized_correct_counts'] = round((merged_df['correct_counts'] / merged_df['total_counts']) * 100, 2)
# Sorting the results to ensure the plot is ordered
merged_df.sort_values(by=['true_continent', 'income_bucket'], ascending=[True, True], inplace=True)
# Plotting with seaborn barplot
plt.figure(figsize=(20, 10))
sns.set_theme(style="whitegrid")
sns.barplot(data=merged_df, x='true_continent', y='normalized_correct_counts', hue='income_bucket', palette='deep')
# Improve the aesthetics
plt.yticks(fontsize=32)
plt.xticks(fontsize=32)
# plt.title('Accuracy across normalized income quartiles', fontsize=24)
plt.xlabel('', fontsize=22)
plt.ylabel('Correct Classification Percentage', fontsize=22)
legend = plt.legend(title='Income Quartile', fontsize=24)
plt.setp(legend.get_title(),fontsize=26)
plt.tight_layout()
# Save the plot
plt.savefig(f"accuracy_by_income_quartiles_{DATA}_street_{MODEL}.png")
# Save the merged dataframe
merged_df.to_csv(f"corrected/accuracy_by_income_quartiles_{DATA}_street_{MODEL}.csv", index=False)
"""#confusion matrix subregion plot
# Load JSON mapping file
with open('unsd_geoscheme.json') as f:
mappings = json.load(f)
# Flatten the JSON mapping for easier lookup
country_to_subregion = {}
subregion_to_continent = {}
for continent, regions in mappings.items():
for subregion, data in regions.items():
subregion_to_continent[subregion] = continent
for country in data['countries_included']:
country_to_subregion[country] = subregion
# Load the CSV file
df = pd.read_csv(PATH)
# Map true subregions from true_country
df['true_sub_region'] = df['true_country'].map(country_to_subregion)
# Replace 'ResponsibleAIPolicyViolation' with 'ResponsibleAI'
df['response'] = df['response'].replace('ResponsibleAIPolicyViolation', 'ResponsibleAI')
# Map predicted subregions
df['predicted_sub_region'] = df['response'].apply(lambda x: x if x in subregion_to_continent else "Invalid")
# Replace invalid responses in the response column
invalids = ["Invalid", "ResponsibleAI"]
df.loc[df['response'].isin(invalids), 'predicted_sub_region'] = df['response']
# Ensure there are no NaNs in true_sub_region or predicted_sub_region
df = df.dropna(subset=['true_sub_region', 'predicted_sub_region'])
# Get the list of all subregions
subregions_set = set(df['true_sub_region'].unique()).union(set(df['predicted_sub_region'].unique()))
# Convert to list and sort by continent, then alphabetically within each continent
sorted_subregions = sorted(subregions_set, key=lambda x: (subregion_to_continent.get(x, ""), x))
# Ensure 'Invalid' and 'ResponsibleAI' are at the end
all_subregions = [subregion for subregion in sorted_subregions if subregion not in invalids] + invalids
# Calculate confusion matrix
conf_matrix = confusion_matrix(df['true_sub_region'], df['predicted_sub_region'], labels=all_subregions)
# Plot confusion matrix
plt.figure(figsize=(20, 15))
h = sns.heatmap(conf_matrix, annot=True, fmt="d", cmap="Blues", xticklabels=all_subregions, yticklabels=all_subregions)
cbar = h.collections[0].colorbar
cbar.ax.tick_params(labelsize=20)
plt.xlabel('Predicted Subregion', fontsize=18)
plt.ylabel('True Subregion', fontsize=18)
plt.xticks(fontsize=20)
plt.yticks(fontsize=20)
# Get the current axes
ax = plt.gca()
# Find the indices for each continent
region_indices = {}
for i, subregion in enumerate(all_subregions):
continent = subregion_to_continent.get(subregion, subregion) # Use subregion name if continent not found
if continent not in region_indices:
region_indices[continent] = []
region_indices[continent].append(i)
# Draw rectangles around each region
for region, indices in region_indices.items():
start = min(indices)
end = max(indices)
width = end - start + 1
rect = Rectangle((start, start), width, width, linewidth=2, edgecolor='black', facecolor='none')
ax.add_patch(rect)
# Add box for 'Invalid' and 'ResponsibleAI'
n = len(all_subregions)
rect_invalid = Rectangle((n-2, n-2), 1, 1, linewidth=2, edgecolor='black', facecolor='none')
rect_responsibleai = Rectangle((n-1, n-1), 1, 1, linewidth=2, edgecolor='black', facecolor='none')
ax.add_patch(rect_invalid)
ax.add_patch(rect_responsibleai)
plt.tight_layout()
# Save the plot
plt.savefig(f"confusion_matrix_subregions_{DATA}_street_{MODEL}.png")
"""
"""#confusion matrix regions
# Load JSON mapping file
with open('unsd_geoscheme.json') as f:
mappings = json.load(f)
# Flatten the JSON mapping for easier lookup
country_to_subregion = {}
subregion_to_continent = {}
for continent, regions in mappings.items():
for subregion, data in regions.items():
subregion_to_continent[subregion] = continent
for country in data['countries_included']:
country_to_subregion[country] = (subregion, continent)
# Load the CSV file
df = pd.read_csv(PATH)
# create a blank column for the predicted continent
df["predicted_continent"] = ""
if MODEL == "llava":
if DATA == "dollar":
invalids = ["Africa", "Americas", "Asia", "Europe", "Oceania", "Antarctica", "Middle East", "ResponsibleAIPolicyViolation"]
if DATA == "dalle":
invalids = ["Africa", "Americas", "Asia", "Europe", "Oceania", "Arctic", "Antarctica", "Middle East", "ResponsibleAIPolicyViolation"]
else:
invalids = ["Africa", "Americas", "Asia", "Europe", "Oceania", "ResponsibleAIPolicyViolation"]
df.loc[df["predicted_continent"] == "", "predicted_continent"] = df.loc[df["predicted_continent"] == "", "response"].map(subregion_to_continent)
for invalid in invalids:
df.loc[df["response"].str.contains(invalid), "predicted_continent"] = invalid
print(df.head(10))
if MODEL == "llava":
if DATA == "dollar":
continent_order = ["Africa", "Americas", "Asia", "Europe", "Oceania", "Antarctica", "Middle East", "ResponsibleAI"]
if DATA == "dalle":
continent_order = ["Africa", "Americas", "Asia", "Europe", "Oceania", "Arctic", "Antarctica", "Middle East", "ResponsibleAI"]
else:
continent_order = ["Africa", "Americas", "Asia", "Europe", "Oceania", "ResponsibleAI"]
conf_matrix = confusion_matrix(df['true_continent'], df['predicted_continent'], labels=invalids)
print(conf_matrix)
# Plot confusion matrix
plt.figure(figsize=(10, 7))
sns.set_theme(style="whitegrid")
sns.heatmap(conf_matrix, annot=True, fmt="d", cmap="Blues", xticklabels=continent_order, yticklabels=continent_order)
plt.xlabel('Predicted Continent')
plt.ylabel('True Continent')
plt.title('Confusion Matrix: True Continent vs Predicted Continent')
plt.tight_layout()
# Save the plot
plt.savefig(f"confusion_matrix_{DATA}_street_{MODEL}.png")
"""
"""#subregion plot
# Load JSON mapping file
with open('unsd_geoscheme.json') as f:
mappings = json.load(f)
# Flatten the JSON mapping for easier lookup
country_to_subregion = {}
for continent, regions in mappings.items():
for subregion, data in regions.items():
for country in data['countries_included']:
country_to_subregion[country] = (subregion, data['continent'])
valid_subregions = set(itertools.chain(*[list(regions.keys()) for continent, regions in mappings.items()]))
# Load the CSV file
df = pd.read_csv(PATH)
# Calculate if the prediction is correct
df['correct_prediction'] = df['response'] == df['true_sub_region']
# Calculate accuracy for each country
accuracy_by_country = df.groupby('true_country')['correct_prediction'].mean() * 100
accuracy_by_country = accuracy_by_country.round(2)
# Convert to DataFrame for better display
accuracy_by_country_df = accuracy_by_country.reset_index()
accuracy_by_country_df.columns = ['true_country', 'accuracy']
# Merge with true_sub_region and true_continent
accuracy_by_country_df = pd.merge(accuracy_by_country_df, df[['true_country', 'true_sub_region', 'true_continent']].drop_duplicates(), on='true_country', how='left')
# first order by true_continent and then within each continent, order by
# true_sub_region and then within each subregion, order by accuracy
accuracy_by_country_df.sort_values(by=['true_continent', 'true_sub_region', 'accuracy'], ascending=[True, True, False], inplace=True)
# Sort by 'true_sub_region' and 'accuracy' for a consistent bar order
# accuracy_by_country_df.sort_values(by=['true_sub_region', 'accuracy'], ascending=[True, False], inplace=True)
# Assign colors to each subregion based on the unique values in 'true_sub_region'
subregion_order = accuracy_by_country_df['true_sub_region'].unique() # Order of subregions for the legend
subregion_colors = sns.color_palette("husl", len(subregion_order)) # Generate as many colors as subregions
subregion_palette = dict(zip(subregion_order, subregion_colors)) # Create a dictionary to map subregions to colors
# Plotting with seaborn barplot
plt.figure(figsize=(20, 10))
sns.set_theme(style="whitegrid")
sns.barplot(data=accuracy_by_country_df, x='true_country', y='accuracy', hue='true_sub_region', dodge=False, palette=subregion_palette)
# Improve the aesthetics
plt.xticks(rotation=90)
plt.xlabel('Country')
plt.ylabel('Accuracy (%)')
plt.legend(title='Subregion')
plt.tight_layout()
# Save the plot and the DataFrame to files
plt.savefig(f"country_wise_accuracy_by_subregion_{DATA}_street_{MODEL}.png")
accuracy_by_country_df.to_csv(f"corrected/country_wise_accuracy_by_subregion_{DATA}_street_{MODEL}.csv", index=False)
"""
"""#continent plot
# Load JSON mapping file
with open('unsd_geoscheme.json') as f:
mappings = json.load(f)
# Flatten the JSON mapping for easier lookup
country_to_subregion = {}
for continent, regions in mappings.items():
for subregion, data in regions.items():
for country in data['countries_included']:
country_to_subregion[country] = (subregion, data['continent'])
valid_subregions = set(itertools.chain(*[list(regions.keys()) for continent, regions in mappings.items()]))
df = pd.read_csv(PATH)
# print(df.head())
# print(df.columns)
# if papua new guinea, print true_continet
# print(df[df['true_country'] == 'Papua New Guinea']['true_continent'])
# Calculate if the prediction is correct
df['correct_prediction'] = df['response'] == df['true_sub_region']
# Calculate accuracy for each country
accuracy_by_country = df.groupby('true_country')['correct_prediction'].mean() * 100
# round to 2 decimal places
accuracy_by_country = accuracy_by_country.round(2)
# Convert to DataFrame for better display
accuracy_by_country_df = accuracy_by_country.reset_index()
accuracy_by_country_df.columns = ['true_country', 'accuracy']
# Merge with true_continent
accuracy_by_country_df = pd.merge(accuracy_by_country_df, df[['true_country', 'true_continent']].drop_duplicates(), on='true_country', how='left')
# Sort by 'true_continent' and 'accuracy' for a consistent bar order
accuracy_by_country_df.sort_values(by=['true_continent', 'accuracy'], ascending=[True, False], inplace=True)
# Assign colors to each continent based on the unique values in 'true_continent'
continent_order = accuracy_by_country_df['true_continent'].unique() # Order of continents for the legend
continent_colors = sns.color_palette("husl", len(continent_order)) # Generate as many colors as continents
continent_palette = dict(zip(continent_order, continent_colors)) # Create a dictionary to map continents to colors
# Plotting with seaborn barplot
plt.figure(figsize=(20, 10))
sns.set_theme(style="whitegrid")
sns.barplot(data=accuracy_by_country_df, x='true_country', y='accuracy', hue='true_continent', dodge=False, palette=continent_palette)
# Improve the aesthetics
plt.xticks(rotation=90)
plt.xlabel('Country')
plt.ylabel('Accuracy (%)')
plt.legend(title='Continent')
plt.tight_layout()
plt.savefig(f"country_wise_accuracy_{DATA}_street_{MODEL}.png")
accuracy_by_country_df.to_csv(f"corrected/country_wise_accuracy_{DATA}_street_{MODEL}.csv", index=False)
"""