Upload data_processing.py
Browse files- data_processing.py +245 -0
data_processing.py
ADDED
@@ -0,0 +1,245 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# -*- coding: utf-8 -*-
|
2 |
+
"""data_processing.ipynb
|
3 |
+
|
4 |
+
Automatically generated by Colab.
|
5 |
+
|
6 |
+
Original file is located at
|
7 |
+
https://colab.research.google.com/drive/1Oz1QL0mD9g3lVBgtmqHa-QiwwIJ2JaX5
|
8 |
+
"""
|
9 |
+
|
10 |
+
import pandas as pd
|
11 |
+
import numpy as np
|
12 |
+
import os
|
13 |
+
from zipfile import ZipFile
|
14 |
+
import re
|
15 |
+
import json
|
16 |
+
import io
|
17 |
+
from PIL import Image, ImageFile
|
18 |
+
ImageFile.LOAD_TRUNCATED_IMAGES = True
|
19 |
+
|
20 |
+
from google.colab import drive
|
21 |
+
drive.mount('/content/drive')
|
22 |
+
|
23 |
+
path = "/content/drive/MyDrive/Duke/aphantasia_drawing_project/"
|
24 |
+
|
25 |
+
data_path = os.path.join(path,"data",'drawing_experiment')
|
26 |
+
|
27 |
+
|
28 |
+
|
29 |
+
df = pd.read_excel(data_path+"/questionnaire-data.xlsx", header=2)
|
30 |
+
|
31 |
+
df["vviq_score"] = np.sum(df.filter(like = "vviq"), axis = 1)
|
32 |
+
df["osiq_score"] = np.sum(df.filter(like = "osiq"), axis = 1)
|
33 |
+
df["treatment"] = np.where(df.vviq_score > 40, "control", "aphantasia")
|
34 |
+
|
35 |
+
df = df.rename(columns={
|
36 |
+
"Sub ID": "sub_id",
|
37 |
+
df.columns[5]: "art_ability",
|
38 |
+
df.columns[6]: "art_experience",
|
39 |
+
df.columns[9]: "difficult",
|
40 |
+
df.columns[10]: "diff_explanation"
|
41 |
+
})
|
42 |
+
|
43 |
+
df.columns = df.columns.str.lower()
|
44 |
+
|
45 |
+
df = df.drop(df.filter(like="unnamed").columns, axis = 1)
|
46 |
+
df = df.drop(df.filter(regex="(vviq|osiq)\d+").columns, axis = 1)
|
47 |
+
|
48 |
+
df[df.columns[df.dtypes == "object"]] = df[df.columns[df.dtypes == "object"]].astype("string")
|
49 |
+
|
50 |
+
df = df.replace([np.nan,pd.NA, "nan","na","NA","n/a","N/A","N/a"], None)
|
51 |
+
|
52 |
+
df.set_index('sub_id', inplace=True)
|
53 |
+
|
54 |
+
|
55 |
+
|
56 |
+
actual_image_path = os.path.join(data_path,"Stimuli","Images")
|
57 |
+
|
58 |
+
|
59 |
+
|
60 |
+
actual_images = {}
|
61 |
+
for image_file in os.listdir(actual_image_path):
|
62 |
+
img_path = os.path.join(actual_image_path, image_file)
|
63 |
+
actual_images[image_file.removesuffix(".jpg")] = Image.open(img_path)
|
64 |
+
|
65 |
+
|
66 |
+
|
67 |
+
key_map = {
|
68 |
+
'high_sun_ajwbpqrwvknlvpeh': 'kitchen',
|
69 |
+
'low_sun_acqsqjhtcbxeomux': 'bedroom',
|
70 |
+
"low_sun_byqgoskwpvsbllvy":"livingroom"
|
71 |
+
}
|
72 |
+
|
73 |
+
for old_key, new_key in key_map.items():
|
74 |
+
actual_images[new_key] = actual_images.pop(old_key)
|
75 |
+
|
76 |
+
aphantasia_drawings_path = os.path.join(data_path,"Drawings","Aphantasia")
|
77 |
+
control_drawings_path = os.path.join(data_path,"Drawings","Control")
|
78 |
+
|
79 |
+
|
80 |
+
|
81 |
+
directories = {
|
82 |
+
"Aphantasia": aphantasia_drawings_path,
|
83 |
+
"Control": control_drawings_path
|
84 |
+
}
|
85 |
+
|
86 |
+
|
87 |
+
|
88 |
+
aphantasia_subs = {i: "Aphantasia" for i in os.listdir(directories["Aphantasia"]) if "sub" in i}
|
89 |
+
control_subs = {i: "Control" for i in os.listdir(directories["Control"]) if "sub" in i}
|
90 |
+
sub_treatment_key = {**aphantasia_subs, **control_subs}
|
91 |
+
|
92 |
+
|
93 |
+
|
94 |
+
def get_sub_files(sub):
|
95 |
+
treatment_group = sub_treatment_key[sub]
|
96 |
+
directory = directories[treatment_group]
|
97 |
+
pattern = re.compile("^.*" + sub + "-[a-z]{3}\d-(kitchen|livingroom|bedroom).*")
|
98 |
+
sub_files = os.listdir(os.path.join(directory, sub))
|
99 |
+
files_needed = {'mem1',"mem2",'mem3','pic1','pic2','pic3'}
|
100 |
+
|
101 |
+
sub_key = {}
|
102 |
+
for f in sub_files:
|
103 |
+
if pattern.match(f):
|
104 |
+
main_path = os.path.join(directory, sub, f)
|
105 |
+
draw_type = f.split("-")[1]
|
106 |
+
label = f.split("-")[2].removesuffix(".jpg")
|
107 |
+
alt_path = os.path.join(directory, sub, "-".join([sub, draw_type]) + ".jpg")
|
108 |
+
try:
|
109 |
+
img = Image.open(main_path)
|
110 |
+
except:
|
111 |
+
img = Image.open(alt_path)
|
112 |
+
sub_key[draw_type] = {
|
113 |
+
"label": label,
|
114 |
+
"drawing": img
|
115 |
+
}
|
116 |
+
|
117 |
+
unknown_drawings = files_needed - sub_key.keys()
|
118 |
+
if unknown_drawings:
|
119 |
+
for unk in unknown_drawings:
|
120 |
+
path = os.path.join(directory, sub, "-".join([sub, unk]) + ".jpg")
|
121 |
+
try:
|
122 |
+
img = Image.open(path)
|
123 |
+
except:
|
124 |
+
img = "No Image"
|
125 |
+
sub_key[unk] = {
|
126 |
+
"label": "unknown",
|
127 |
+
"drawing": img
|
128 |
+
}
|
129 |
+
return sub_key
|
130 |
+
|
131 |
+
|
132 |
+
|
133 |
+
|
134 |
+
|
135 |
+
subject_data = {}
|
136 |
+
for sub in iter(sub_treatment_key):
|
137 |
+
subject_data[sub] = get_sub_files(sub)
|
138 |
+
|
139 |
+
def is_image_blank(image):
|
140 |
+
if image.mode != 'RGB':
|
141 |
+
image = image.convert('RGB')
|
142 |
+
pixels = list(image.getdata())
|
143 |
+
return all(pixel == (255, 255, 255) for pixel in pixels)
|
144 |
+
|
145 |
+
for sub in iter(subject_data):
|
146 |
+
dat = subject_data[sub]
|
147 |
+
for key in dat.keys():
|
148 |
+
if is_image_blank(dat[key]["drawing"]):
|
149 |
+
dat[key]["label"] = "blank"
|
150 |
+
|
151 |
+
|
152 |
+
|
153 |
+
subs_missing_labels = {}
|
154 |
+
for sub in iter(subject_data):
|
155 |
+
dat = subject_data[sub]
|
156 |
+
for key in dat.keys():
|
157 |
+
if "unknown" in dat[key]["label"]:
|
158 |
+
if sub not in subs_missing_labels:
|
159 |
+
subs_missing_labels[sub] = []
|
160 |
+
subs_missing_labels[sub].append(key)
|
161 |
+
|
162 |
+
subs_missing_labels
|
163 |
+
|
164 |
+
subject_data["sub8"]["pic3"]["label"] = "livingroom"
|
165 |
+
subject_data["sub6"]["pic3"]["label"] = "bedroom"
|
166 |
+
subject_data["sub6"]["pic1"]["label"] = "kitchen"
|
167 |
+
|
168 |
+
|
169 |
+
|
170 |
+
def clean_sub_dat(sub):
|
171 |
+
id = int(sub[3:])
|
172 |
+
treatment_group = sub_treatment_key[sub]
|
173 |
+
if id in df.index:
|
174 |
+
demographics_dict = df.loc[id].to_dict()
|
175 |
+
else:
|
176 |
+
demographics_dict = {}
|
177 |
+
demographics_dict.pop("treatment",None)
|
178 |
+
drawings = {
|
179 |
+
"bedroom": {},
|
180 |
+
"kitchen": {},
|
181 |
+
"livingroom": {}
|
182 |
+
}
|
183 |
+
for draw_type, draw_data in subject_data[sub].items():
|
184 |
+
t = "memory" if draw_type[:-1] == "mem" else "perception"
|
185 |
+
for d in drawings.keys():
|
186 |
+
if draw_data["label"] == d:
|
187 |
+
drawings[d][t] = draw_data["drawing"]
|
188 |
+
|
189 |
+
return {
|
190 |
+
"subject_id": id,
|
191 |
+
"treatment": treatment_group,
|
192 |
+
"demographics": demographics_dict,
|
193 |
+
"drawings": drawings,
|
194 |
+
"image": actual_images
|
195 |
+
}
|
196 |
+
|
197 |
+
|
198 |
+
|
199 |
+
full_data = []
|
200 |
+
for s in subject_data.keys():
|
201 |
+
full_data.append(clean_sub_dat(s))
|
202 |
+
|
203 |
+
"""160,161,162 removed, they dont have images"""
|
204 |
+
|
205 |
+
|
206 |
+
|
207 |
+
full_df = pd.json_normalize(full_data)
|
208 |
+
|
209 |
+
|
210 |
+
|
211 |
+
def image_to_byt(img, size=(224, 224)):
|
212 |
+
if pd.isna(img):
|
213 |
+
return None
|
214 |
+
img_resized = img.resize(size)
|
215 |
+
img_byte_arr = io.BytesIO()
|
216 |
+
img_resized.save(img_byte_arr, format='PNG')
|
217 |
+
return img_byte_arr.getvalue()
|
218 |
+
|
219 |
+
drawing_columns = [col for col in full_df.columns if "drawings" in col or "image" in col]
|
220 |
+
|
221 |
+
|
222 |
+
|
223 |
+
for col in drawing_columns:
|
224 |
+
full_df[col] = full_df[col].apply(image_to_byt)
|
225 |
+
|
226 |
+
def safe_convert_to_int(value):
|
227 |
+
try:
|
228 |
+
return int(value)
|
229 |
+
except (ValueError, TypeError):
|
230 |
+
return -99
|
231 |
+
|
232 |
+
col_to_process = [
|
233 |
+
"demographics.age",
|
234 |
+
"demographics.art_ability",
|
235 |
+
"demographics.vviq_score",
|
236 |
+
"demographics.osiq_score"
|
237 |
+
]
|
238 |
+
|
239 |
+
for col in col_to_process:
|
240 |
+
full_df[col] = full_df[col].apply(safe_convert_to_int)
|
241 |
+
|
242 |
+
full_data_path = os.path.join(path, "data","aphantasia_data.parquet")
|
243 |
+
|
244 |
+
full_df.to_parquet(full_data_path, index=False)
|
245 |
+
|