File size: 12,383 Bytes
9313bc1 ee44327 9313bc1 bf0e4cf 9313bc1 ee44327 9313bc1 a942e66 9313bc1 a942e66 9313bc1 a942e66 9313bc1 ee44327 9313bc1 d901742 9313bc1 d901742 9313bc1 d901742 9313bc1 d901742 9313bc1 ee44327 9313bc1 ee44327 9313bc1 ee44327 9313bc1 ee44327 9313bc1 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 |
import json
import os
import pickle
import random
import time
from collections import Counter
from datetime import datetime
from glob import glob
import gdown
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import seaborn as sns
import streamlit as st
from PIL import Image
import SessionState
from download_utils import *
from image_utils import *
random.seed(datetime.now())
np.random.seed(int(time.time()))
NUMBER_OF_TRIALS = 20
CLASSIFIER_TAG = ""
explaination_functions = [load_chm_nns, load_knn_nns]
selected_xai_tool = None
# Config
folder_to_name = {}
class_descriptions = {}
classifier_predictions = {}
selected_dataset = "Final"
root_visualization_dir = "./visualizations/"
viz_url = "https://static.taesiri.com/xai/Final.zip"
viz_archivefile = "Final.zip"
demonstration_url = "https://static.taesiri.com/xai/demonstrations.zip"
demonst_zipfile = "demonstrations.zip"
picklefile_url = "https://static.taesiri.com/xai/Task1_Results_CHM_and_EMD.pickle"
prediction_root = "./predictions/"
prediction_pickle = f"{prediction_root}predictions.pickle"
# Get the Data
download_files(
root_visualization_dir,
viz_url,
viz_archivefile,
demonstration_url,
demonst_zipfile,
picklefile_url,
prediction_root,
prediction_pickle,
)
################################################
# GLOBAL VARIABLES
app_mode = ""
## Shared/Global Information
with open("imagenet-labels.json", "rb") as f:
folder_to_name = json.load(f)
with open("gloss.txt", "r") as f:
description_file = f.readlines()
class_descriptions = {l.split("\t")[0]: l.split("\t")[1] for l in description_file}
################################################
with open(prediction_pickle, "rb") as f:
classifier_predictions = pickle.load(f)
# SESSION STATE
session_state = SessionState.get(
page=1,
first_run=1,
user_feedback={},
queries=[],
is_classifier_correct={},
XAI_tool="Unselected",
)
################################################
def resmaple_queries():
if session_state.first_run == 1:
both_correct = glob(
root_visualization_dir + selected_dataset + "/Both_correct/*.JPEG"
)
both_wrong = glob(
root_visualization_dir + selected_dataset + "/Both_wrong/*.JPEG"
)
correct_samples = list(
np.random.choice(a=both_correct, size=NUMBER_OF_TRIALS // 2, replace=False)
)
wrong_samples = list(
np.random.choice(a=both_wrong, size=NUMBER_OF_TRIALS // 2, replace=False)
)
all_images = correct_samples + wrong_samples
random.shuffle(all_images)
session_state.queries = all_images
session_state.first_run = -1
# RESET INTERACTIONS
session_state.user_feedback = {}
session_state.is_classifier_correct = {}
def render_experiment(query):
current_query = session_state.queries[query]
query_id = os.path.basename(current_query)
predicted_wnid = classifier_predictions[query_id][f"{CLASSIFIER_TAG}-predictions"]
prediction_confidence = classifier_predictions[query_id][
f"{CLASSIFIER_TAG}-confidence"
]
prediction_label = folder_to_name[predicted_wnid]
class_def = class_descriptions[predicted_wnid]
session_state.is_classifier_correct[query_id] = classifier_predictions[query_id][
f"{CLASSIFIER_TAG.upper()}-Output"
]
################################### SHOW QUERY and PREDICTION
col1, col2 = st.columns(2)
with col1:
st.image(load_query(current_query), caption=f"Query ID: {query_id}")
with col2:
################################### SHOW DESCRIPTION OF CLASS
with st.expander("Show Class Description"):
st.write(f"**Name**: {prediction_label}")
st.write("**Class Definition**:")
st.markdown("`" + class_def + "`")
st.image(
Image.open(f"demonstrations/{predicted_wnid}.jpeg"),
caption=f"Class Explanation",
use_column_width=True,
)
default_value = 0
if query_id in session_state.user_feedback.keys():
if session_state.user_feedback[query_id] == "Correct":
default_value = 1
elif session_state.user_feedback[query_id] == "Wrong":
default_value = 2
session_state.user_feedback[query_id] = st.radio(
"What do you think about model's prediction?",
("-", "Correct", "Wrong"),
key=query_id,
index=default_value,
)
st.write(f"**Model Prediction**: {prediction_label}")
st.write(f"**Model Confidence**: {prediction_confidence}")
################################### SHOW Model Explanation
if selected_xai_tool is not None:
st.image(
selected_xai_tool(current_query),
caption=f"Explaination",
use_column_width=True,
)
################################### SHOW DEBUG INFO
if st.button("Debug: Show Everything"):
st.image(Image.open(current_query))
def render_results():
user_correct_guess = 0
for q in session_state.user_feedback.keys():
uf = True if session_state.user_feedback[q] == 'Correct' else False
if session_state.is_classifier_correct[q] == uf:
user_correct_guess += 1
st.write(
f"User performance on {CLASSIFIER_TAG}: {user_correct_guess} out of {len( session_state.user_feedback)} Correct"
)
st.markdown("## User Performance Breakdown")
categories = [
"Correct",
"Wrong",
] # set(session_state.is_classifier_correct.values())
breakdown_stats_correct = {c: 0 for c in categories}
breakdown_stats_wrong = {c: 0 for c in categories}
experiment_summary = []
for q in session_state.user_feedback.keys():
category = "Correct" if session_state.is_classifier_correct[q] else "Wrong"
is_user_correct = category == session_state.user_feedback[q]
if is_user_correct:
breakdown_stats_correct[category] += 1
else:
breakdown_stats_wrong[category] += 1
experiment_summary.append(
[
q,
classifier_predictions[q]["real-gts"],
folder_to_name[
classifier_predictions[q][f"{CLASSIFIER_TAG}-predictions"]
],
category,
session_state.user_feedback[q],
is_user_correct,
]
)
################################### Summary Table
experiment_summary_df = pd.DataFrame.from_records(
experiment_summary,
columns=[
"Query",
"GT Labels",
f"{CLASSIFIER_TAG} Prediction",
"Category",
"User Prediction",
"Is User Prediction Correct",
],
)
st.write("Summary", experiment_summary_df)
csv = convert_df(experiment_summary_df)
st.download_button(
"Press to Download", csv, "summary.csv", "text/csv", key="download-records"
)
################################### SHOW BREAKDOWN
user_pf_by_model_pred = experiment_summary_df.groupby("Category").agg(
{"Is User Prediction Correct": ["count", "sum", "mean"]}
)
# rename columns
user_pf_by_model_pred.columns = user_pf_by_model_pred.columns.droplevel(0)
user_pf_by_model_pred.columns = [
"Count",
"Correct User Guess",
"Mean User Performance",
]
user_pf_by_model_pred.index.name = "Model Prediction"
st.write("User performance break down by Model prediction:", user_pf_by_model_pred)
csv = convert_df(user_pf_by_model_pred)
st.download_button(
"Press to Download",
csv,
"user-performance-by-model-prediction.csv",
"text/csv",
key="download-performance-by-model-prediction",
)
################################### CONFUSION MATRIX
confusion_matrix = pd.crosstab(
experiment_summary_df["Category"],
experiment_summary_df["User Prediction"],
rownames=["Actual"],
colnames=["Predicted"],
)
st.write("Confusion Matrix", confusion_matrix)
csv = convert_df(confusion_matrix)
st.download_button(
"Press to Download",
csv,
"confusion-matrix.csv",
"text/csv",
key="download-confusiion-matrix",
)
def render_menu():
# Render the readme as markdown using st.markdown.
readme_text = st.markdown(
"""
# Instructions
```
When testing this study, you should first see the class definition, then hide the expander and see the query.
```
"""
)
app_mode = st.selectbox(
"Choose the page to show:",
["Experiment Instruction", "Start Experiment", "See the Results"],
)
if app_mode == "Experiment Instruction":
st.success("To continue select an option in the dropdown menu.")
elif app_mode == "Start Experiment":
# Clear Canvas
readme_text.empty()
page_id = session_state.page
col1, col4, col2, col3 = st.columns(4)
prev_page = col1.button("Previous Image")
if prev_page:
page_id -= 1
if page_id < 1:
page_id = 1
next_page = col2.button("Next Image")
if next_page:
page_id += 1
if page_id > NUMBER_OF_TRIALS:
page_id = NUMBER_OF_TRIALS
if page_id == NUMBER_OF_TRIALS:
st.success(
'You have reached the last image. Please go to the "Results" page to see your performance.'
)
if st.button("View"):
app_mode = "See the Results"
if col3.button("Resample"):
st.write("Restarting ...")
page_id = 1
session_state.first_run = 1
resmaple_queries()
session_state.page = page_id
st.write(f"Render Experiment: {session_state.page}")
render_experiment(session_state.page - 1)
elif app_mode == "See the Results":
readme_text.empty()
st.write("Results Summary")
render_results()
def main():
global app_mode
global session_state
global selected_xai_tool
global CLASSIFIER_TAG
# Set the session state
# State Management and General Setup
st.set_page_config(layout="wide")
st.title("TASK - 1 - ImageNetREAL")
# st.write(classifier_predictions.keys())
# st.write(classifier_predictions["ILSVRC2012_val_00024646.JPEG"])
options = [
"Unselected",
"NOXAI",
"KNN",
"EMD Nearest Neighbors",
"EMD Correspondence",
"CHM Nearest Neighbors",
"CHM Correspondence",
]
st.markdown(
""" <style>
div[role="radiogroup"] > :first-child{
display: none !important;
}
</style>
""",
unsafe_allow_html=True,
)
if session_state.XAI_tool == "Unselected":
default = options.index(session_state.XAI_tool)
session_state.XAI_tool = st.radio(
"What explaination tool do you want to evaluate?",
options,
key="which_xai",
index=default,
)
# print(session_state.XAI_tool)
if session_state.XAI_tool != "Unselected":
st.markdown(f"## SELECTED METHOD ``{session_state.XAI_tool}``")
if session_state.XAI_tool == "NOXAI":
CLASSIFIER_TAG = "knn"
selected_xai_tool = None
elif session_state.XAI_tool == "KNN":
selected_xai_tool = load_knn_nns
CLASSIFIER_TAG = "knn"
elif session_state.XAI_tool == "CHM Nearest Neighbors":
selected_xai_tool = load_chm_nns
CLASSIFIER_TAG = "CHM"
elif session_state.XAI_tool == "CHM Correspondence":
selected_xai_tool = load_chm_corrs
CLASSIFIER_TAG = "CHM"
elif session_state.XAI_tool == "EMD Nearest Neighbors":
selected_xai_tool = load_emd_nns
CLASSIFIER_TAG = "EMD"
elif session_state.XAI_tool == "EMD Correspondence":
selected_xai_tool = load_emd_corrs
CLASSIFIER_TAG = "EMD"
resmaple_queries()
render_menu()
if __name__ == "__main__":
main()
|