Spaces:
Running
Running
Joshua Lochner
commited on
Commit
·
dccb47b
1
Parent(s):
67d0193
Separate missing and incorrect detection logic
Browse files- src/evaluate.py +140 -78
src/evaluate.py
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
|
| 2 |
from model import get_model_tokenizer_classifier, InferenceArguments
|
| 3 |
-
from utils import jaccard
|
| 4 |
from transformers import HfArgumentParser
|
| 5 |
-
from preprocess import get_words
|
| 6 |
from shared import GeneralArguments, DatasetArguments
|
| 7 |
from predict import predict
|
| 8 |
from segment import extract_segment, word_start, word_end, SegmentationArguments, add_labels_to_words
|
|
@@ -31,6 +31,19 @@ class EvaluationArguments(InferenceArguments):
|
|
| 31 |
}
|
| 32 |
)
|
| 33 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 34 |
|
| 35 |
def attach_predictions_to_sponsor_segments(predictions, sponsor_segments):
|
| 36 |
"""Attach sponsor segments to closest prediction"""
|
|
@@ -46,7 +59,7 @@ def attach_predictions_to_sponsor_segments(predictions, sponsor_segments):
|
|
| 46 |
prediction['best_overlap'] = j
|
| 47 |
prediction['best_sponsorship'] = sponsor_segment
|
| 48 |
|
| 49 |
-
|
| 50 |
|
| 51 |
|
| 52 |
def calculate_metrics(labelled_words, predictions):
|
|
@@ -130,6 +143,10 @@ def main():
|
|
| 130 |
|
| 131 |
evaluation_args, dataset_args, segmentation_args, general_args = hf_parser.parse_args_into_dataclasses()
|
| 132 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 133 |
# Load labelled data:
|
| 134 |
final_path = os.path.join(
|
| 135 |
dataset_args.data_dir, dataset_args.processed_file)
|
|
@@ -158,14 +175,22 @@ def main():
|
|
| 158 |
if evaluation_args.max_videos is not None:
|
| 159 |
video_ids = video_ids[:evaluation_args.max_videos]
|
| 160 |
|
| 161 |
-
|
| 162 |
|
| 163 |
-
|
| 164 |
-
|
| 165 |
-
|
| 166 |
-
|
|
|
|
|
|
|
| 167 |
|
| 168 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 169 |
|
| 170 |
try:
|
| 171 |
with tqdm(video_ids) as progress:
|
|
@@ -176,53 +201,77 @@ def main():
|
|
| 176 |
if not words:
|
| 177 |
continue
|
| 178 |
|
| 179 |
-
# Make predictions
|
| 180 |
-
predictions = predict(video_id, model, tokenizer, segmentation_args,
|
| 181 |
-
classifier=classifier,
|
| 182 |
-
min_probability=evaluation_args.min_probability)
|
| 183 |
-
|
| 184 |
# Get labels
|
| 185 |
sponsor_segments = final_data.get(video_id)
|
| 186 |
-
if sponsor_segments:
|
| 187 |
-
labelled_words = add_labels_to_words(
|
| 188 |
-
words, sponsor_segments)
|
| 189 |
-
met = calculate_metrics(labelled_words, predictions)
|
| 190 |
-
met['video_id'] = video_id
|
| 191 |
|
| 192 |
-
|
|
|
|
|
|
|
| 193 |
|
| 194 |
-
|
| 195 |
-
|
| 196 |
-
|
| 197 |
-
|
| 198 |
|
| 199 |
-
|
| 200 |
-
|
| 201 |
-
|
| 202 |
-
|
| 203 |
-
'f-score': total_fscore/len(out_metrics)
|
| 204 |
-
})
|
| 205 |
|
| 206 |
-
|
| 207 |
-
|
|
|
|
| 208 |
|
| 209 |
-
|
| 210 |
-
|
| 211 |
-
prediction for prediction in predictions if prediction['best_sponsorship'] is None]
|
| 212 |
|
| 213 |
-
|
| 214 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 215 |
|
| 216 |
segments_to_check = []
|
| 217 |
texts = [] # Texts to send through tokenizer
|
| 218 |
for sponsor_segment in sponsor_segments:
|
| 219 |
segment_words = extract_segment(
|
| 220 |
words, sponsor_segment['start'], sponsor_segment['end'])
|
| 221 |
-
sponsor_segment['text'] = ' '.join(
|
| 222 |
-
|
| 223 |
|
| 224 |
-
duration = sponsor_segment['end'] -
|
| 225 |
-
|
|
|
|
|
|
|
| 226 |
if wps < 1.5:
|
| 227 |
continue
|
| 228 |
|
|
@@ -231,18 +280,24 @@ def main():
|
|
| 231 |
if sponsor_segment['locked']:
|
| 232 |
continue
|
| 233 |
|
|
|
|
|
|
|
| 234 |
texts.append(sponsor_segment['cleaned_text'])
|
| 235 |
segments_to_check.append(sponsor_segment)
|
| 236 |
|
| 237 |
-
if segments_to_check: #
|
| 238 |
|
| 239 |
segments_scores = classifier(texts)
|
| 240 |
|
|
|
|
| 241 |
for segment, scores in zip(segments_to_check, segments_scores):
|
|
|
|
|
|
|
| 242 |
prediction = max(scores, key=lambda x: x['score'])
|
| 243 |
predicted_category = prediction['label'].lower()
|
| 244 |
|
| 245 |
if predicted_category == segment['category']:
|
|
|
|
| 246 |
continue # Ignore correct segments
|
| 247 |
|
| 248 |
segment.update({
|
|
@@ -252,18 +307,19 @@ def main():
|
|
| 252 |
|
| 253 |
incorrect_segments.append(segment)
|
| 254 |
|
| 255 |
-
|
| 256 |
-
|
| 257 |
-
|
| 258 |
-
|
| 259 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 260 |
|
| 261 |
if missed_segments or incorrect_segments:
|
| 262 |
-
for z in missed_segments:
|
| 263 |
-
# Attach original text to missed segments
|
| 264 |
-
# (Already added to incorrect segments)
|
| 265 |
-
z['text'] = ' '.join(x['text']
|
| 266 |
-
for x in z.pop('words', []))
|
| 267 |
|
| 268 |
if evaluation_args.output_as_json:
|
| 269 |
to_print = {'video_id': video_id}
|
|
@@ -274,23 +330,25 @@ def main():
|
|
| 274 |
if incorrect_segments:
|
| 275 |
to_print['incorrect'] = incorrect_segments
|
| 276 |
|
| 277 |
-
|
|
|
|
| 278 |
else:
|
| 279 |
-
|
| 280 |
f'Issues identified for {video_id} (#{video_index})')
|
| 281 |
# Potentially missed segments (model predicted, but not in database)
|
| 282 |
if missed_segments:
|
| 283 |
-
|
| 284 |
segments_to_submit = []
|
| 285 |
for i, missed_segment in enumerate(missed_segments, start=1):
|
| 286 |
-
|
| 287 |
missed_segment['start']), '-->', seconds_to_time(missed_segment['end']))
|
| 288 |
-
|
| 289 |
-
|
| 290 |
-
|
|
|
|
| 291 |
if 'probability' in missed_segment:
|
| 292 |
-
|
| 293 |
-
|
| 294 |
|
| 295 |
segments_to_submit.append({
|
| 296 |
'segment': [missed_segment['start'], missed_segment['end']],
|
|
@@ -299,33 +357,37 @@ def main():
|
|
| 299 |
})
|
| 300 |
|
| 301 |
json_data = quote(json.dumps(segments_to_submit))
|
| 302 |
-
|
| 303 |
f'\tSubmit: https://www.youtube.com/watch?v={video_id}#segments={json_data}')
|
| 304 |
|
| 305 |
# Incorrect segments (in database, but incorrectly classified)
|
| 306 |
if incorrect_segments:
|
| 307 |
-
|
| 308 |
for i, incorrect_segment in enumerate(incorrect_segments, start=1):
|
| 309 |
-
|
| 310 |
incorrect_segment['start']), '-->', seconds_to_time(incorrect_segment['end']))
|
| 311 |
|
| 312 |
-
|
| 313 |
-
|
| 314 |
-
|
| 315 |
-
|
| 316 |
-
|
| 317 |
-
|
| 318 |
-
|
| 319 |
-
|
| 320 |
-
|
| 321 |
-
|
| 322 |
-
|
| 323 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 324 |
for item in incorrect_segment['scores']:
|
| 325 |
-
|
| 326 |
f"\t\t\t{item['label']}: {item['score']}")
|
| 327 |
|
| 328 |
-
|
| 329 |
|
| 330 |
except KeyboardInterrupt:
|
| 331 |
pass
|
|
|
|
| 1 |
|
| 2 |
from model import get_model_tokenizer_classifier, InferenceArguments
|
| 3 |
+
from utils import jaccard, safe_print
|
| 4 |
from transformers import HfArgumentParser
|
| 5 |
+
from preprocess import get_words, clean_text
|
| 6 |
from shared import GeneralArguments, DatasetArguments
|
| 7 |
from predict import predict
|
| 8 |
from segment import extract_segment, word_start, word_end, SegmentationArguments, add_labels_to_words
|
|
|
|
| 31 |
}
|
| 32 |
)
|
| 33 |
|
| 34 |
+
skip_missing: bool = field(
|
| 35 |
+
default=False,
|
| 36 |
+
metadata={
|
| 37 |
+
'help': 'Whether to skip checking for missing segments. If False, predictions will be made.'
|
| 38 |
+
}
|
| 39 |
+
)
|
| 40 |
+
skip_incorrect: bool = field(
|
| 41 |
+
default=False,
|
| 42 |
+
metadata={
|
| 43 |
+
'help': 'Whether to skip checking for incorrect segments. If False, classifications will be made on existing segments.'
|
| 44 |
+
}
|
| 45 |
+
)
|
| 46 |
+
|
| 47 |
|
| 48 |
def attach_predictions_to_sponsor_segments(predictions, sponsor_segments):
|
| 49 |
"""Attach sponsor segments to closest prediction"""
|
|
|
|
| 59 |
prediction['best_overlap'] = j
|
| 60 |
prediction['best_sponsorship'] = sponsor_segment
|
| 61 |
|
| 62 |
+
return sponsor_segments
|
| 63 |
|
| 64 |
|
| 65 |
def calculate_metrics(labelled_words, predictions):
|
|
|
|
| 143 |
|
| 144 |
evaluation_args, dataset_args, segmentation_args, general_args = hf_parser.parse_args_into_dataclasses()
|
| 145 |
|
| 146 |
+
if evaluation_args.skip_missing and evaluation_args.skip_incorrect:
|
| 147 |
+
logger.error('ERROR: Nothing to do')
|
| 148 |
+
return
|
| 149 |
+
|
| 150 |
# Load labelled data:
|
| 151 |
final_path = os.path.join(
|
| 152 |
dataset_args.data_dir, dataset_args.processed_file)
|
|
|
|
| 175 |
if evaluation_args.max_videos is not None:
|
| 176 |
video_ids = video_ids[:evaluation_args.max_videos]
|
| 177 |
|
| 178 |
+
out_metrics = []
|
| 179 |
|
| 180 |
+
all_metrics = {}
|
| 181 |
+
if not evaluation_args.skip_missing:
|
| 182 |
+
all_metrics['total_prediction_accuracy'] = 0
|
| 183 |
+
all_metrics['total_prediction_precision'] = 0
|
| 184 |
+
all_metrics['total_prediction_recall'] = 0
|
| 185 |
+
all_metrics['total_prediction_fscore'] = 0
|
| 186 |
|
| 187 |
+
if not evaluation_args.skip_incorrect:
|
| 188 |
+
all_metrics['classifier_segment_correct'] = 0
|
| 189 |
+
all_metrics['classifier_segment_count'] = 0
|
| 190 |
+
|
| 191 |
+
metric_count = 0
|
| 192 |
+
|
| 193 |
+
postfix_info = {}
|
| 194 |
|
| 195 |
try:
|
| 196 |
with tqdm(video_ids) as progress:
|
|
|
|
| 201 |
if not words:
|
| 202 |
continue
|
| 203 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 204 |
# Get labels
|
| 205 |
sponsor_segments = final_data.get(video_id)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 206 |
|
| 207 |
+
# Reset previous
|
| 208 |
+
missed_segments = []
|
| 209 |
+
incorrect_segments = []
|
| 210 |
|
| 211 |
+
current_metrics = {
|
| 212 |
+
'video_id': video_id
|
| 213 |
+
}
|
| 214 |
+
metric_count += 1
|
| 215 |
|
| 216 |
+
if not evaluation_args.skip_missing: # Make predictions
|
| 217 |
+
predictions = predict(video_id, model, tokenizer, segmentation_args,
|
| 218 |
+
classifier=classifier,
|
| 219 |
+
min_probability=evaluation_args.min_probability)
|
|
|
|
|
|
|
| 220 |
|
| 221 |
+
if sponsor_segments:
|
| 222 |
+
labelled_words = add_labels_to_words(
|
| 223 |
+
words, sponsor_segments)
|
| 224 |
|
| 225 |
+
current_metrics.update(
|
| 226 |
+
calculate_metrics(labelled_words, predictions))
|
|
|
|
| 227 |
|
| 228 |
+
all_metrics['total_prediction_accuracy'] += current_metrics['accuracy']
|
| 229 |
+
all_metrics['total_prediction_precision'] += current_metrics['precision']
|
| 230 |
+
all_metrics['total_prediction_recall'] += current_metrics['recall']
|
| 231 |
+
all_metrics['total_prediction_fscore'] += current_metrics['f-score']
|
| 232 |
+
|
| 233 |
+
# Just for display purposes
|
| 234 |
+
postfix_info.update({
|
| 235 |
+
'accuracy': all_metrics['total_prediction_accuracy']/metric_count,
|
| 236 |
+
'precision': all_metrics['total_prediction_precision']/metric_count,
|
| 237 |
+
'recall': all_metrics['total_prediction_recall']/metric_count,
|
| 238 |
+
'f-score': all_metrics['total_prediction_fscore']/metric_count,
|
| 239 |
+
})
|
| 240 |
+
|
| 241 |
+
sponsor_segments = attach_predictions_to_sponsor_segments(
|
| 242 |
+
predictions, sponsor_segments)
|
| 243 |
+
|
| 244 |
+
# Identify possible issues:
|
| 245 |
+
for prediction in predictions:
|
| 246 |
+
if prediction['best_sponsorship'] is not None:
|
| 247 |
+
continue
|
| 248 |
+
|
| 249 |
+
prediction_words = prediction.pop('words', [])
|
| 250 |
+
|
| 251 |
+
# Attach original text to missed segments
|
| 252 |
+
prediction['text'] = ' '.join(
|
| 253 |
+
x['text'] for x in prediction_words)
|
| 254 |
+
missed_segments.append(prediction)
|
| 255 |
+
|
| 256 |
+
else:
|
| 257 |
+
# Not in database (all segments missed)
|
| 258 |
+
missed_segments = predictions
|
| 259 |
+
|
| 260 |
+
if not evaluation_args.skip_incorrect and sponsor_segments:
|
| 261 |
+
# Check for incorrect segments using the classifier
|
| 262 |
|
| 263 |
segments_to_check = []
|
| 264 |
texts = [] # Texts to send through tokenizer
|
| 265 |
for sponsor_segment in sponsor_segments:
|
| 266 |
segment_words = extract_segment(
|
| 267 |
words, sponsor_segment['start'], sponsor_segment['end'])
|
| 268 |
+
sponsor_segment['text'] = ' '.join(
|
| 269 |
+
x['text'] for x in segment_words)
|
| 270 |
|
| 271 |
+
duration = sponsor_segment['end'] - \
|
| 272 |
+
sponsor_segment['start']
|
| 273 |
+
wps = (len(segment_words) /
|
| 274 |
+
duration) if duration > 0 else 0
|
| 275 |
if wps < 1.5:
|
| 276 |
continue
|
| 277 |
|
|
|
|
| 280 |
if sponsor_segment['locked']:
|
| 281 |
continue
|
| 282 |
|
| 283 |
+
sponsor_segment['cleaned_text'] = clean_text(
|
| 284 |
+
sponsor_segment['text'])
|
| 285 |
texts.append(sponsor_segment['cleaned_text'])
|
| 286 |
segments_to_check.append(sponsor_segment)
|
| 287 |
|
| 288 |
+
if segments_to_check: # Some segments to check
|
| 289 |
|
| 290 |
segments_scores = classifier(texts)
|
| 291 |
|
| 292 |
+
num_correct = 0
|
| 293 |
for segment, scores in zip(segments_to_check, segments_scores):
|
| 294 |
+
all_metrics['classifier_segment_count'] += 1
|
| 295 |
+
|
| 296 |
prediction = max(scores, key=lambda x: x['score'])
|
| 297 |
predicted_category = prediction['label'].lower()
|
| 298 |
|
| 299 |
if predicted_category == segment['category']:
|
| 300 |
+
num_correct += 1
|
| 301 |
continue # Ignore correct segments
|
| 302 |
|
| 303 |
segment.update({
|
|
|
|
| 307 |
|
| 308 |
incorrect_segments.append(segment)
|
| 309 |
|
| 310 |
+
current_metrics['num_segments'] = len(
|
| 311 |
+
segments_to_check)
|
| 312 |
+
current_metrics['classified_correct'] = num_correct
|
| 313 |
+
|
| 314 |
+
all_metrics['classifier_segment_correct'] += num_correct
|
| 315 |
+
|
| 316 |
+
postfix_info['classifier_accuracy'] = all_metrics['classifier_segment_correct'] / \
|
| 317 |
+
all_metrics['classifier_segment_count']
|
| 318 |
+
|
| 319 |
+
out_metrics.append(current_metrics)
|
| 320 |
+
progress.set_postfix(postfix_info)
|
| 321 |
|
| 322 |
if missed_segments or incorrect_segments:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 323 |
|
| 324 |
if evaluation_args.output_as_json:
|
| 325 |
to_print = {'video_id': video_id}
|
|
|
|
| 330 |
if incorrect_segments:
|
| 331 |
to_print['incorrect'] = incorrect_segments
|
| 332 |
|
| 333 |
+
safe_print(json.dumps(to_print))
|
| 334 |
+
|
| 335 |
else:
|
| 336 |
+
safe_print(
|
| 337 |
f'Issues identified for {video_id} (#{video_index})')
|
| 338 |
# Potentially missed segments (model predicted, but not in database)
|
| 339 |
if missed_segments:
|
| 340 |
+
safe_print(' - Missed segments:')
|
| 341 |
segments_to_submit = []
|
| 342 |
for i, missed_segment in enumerate(missed_segments, start=1):
|
| 343 |
+
safe_print(f'\t#{i}:', seconds_to_time(
|
| 344 |
missed_segment['start']), '-->', seconds_to_time(missed_segment['end']))
|
| 345 |
+
safe_print('\t\tText: "',
|
| 346 |
+
missed_segment['text'], '"', sep='')
|
| 347 |
+
safe_print('\t\tCategory:',
|
| 348 |
+
missed_segment.get('category'))
|
| 349 |
if 'probability' in missed_segment:
|
| 350 |
+
safe_print('\t\tProbability:',
|
| 351 |
+
missed_segment['probability'])
|
| 352 |
|
| 353 |
segments_to_submit.append({
|
| 354 |
'segment': [missed_segment['start'], missed_segment['end']],
|
|
|
|
| 357 |
})
|
| 358 |
|
| 359 |
json_data = quote(json.dumps(segments_to_submit))
|
| 360 |
+
safe_print(
|
| 361 |
f'\tSubmit: https://www.youtube.com/watch?v={video_id}#segments={json_data}')
|
| 362 |
|
| 363 |
# Incorrect segments (in database, but incorrectly classified)
|
| 364 |
if incorrect_segments:
|
| 365 |
+
safe_print(' - Incorrect segments:')
|
| 366 |
for i, incorrect_segment in enumerate(incorrect_segments, start=1):
|
| 367 |
+
safe_print(f'\t#{i}:', seconds_to_time(
|
| 368 |
incorrect_segment['start']), '-->', seconds_to_time(incorrect_segment['end']))
|
| 369 |
|
| 370 |
+
safe_print(
|
| 371 |
+
'\t\tText: "', incorrect_segment['text'], '"', sep='')
|
| 372 |
+
safe_print(
|
| 373 |
+
'\t\tUUID:', incorrect_segment['uuid'])
|
| 374 |
+
safe_print(
|
| 375 |
+
'\t\tVotes:', incorrect_segment['votes'])
|
| 376 |
+
safe_print(
|
| 377 |
+
'\t\tViews:', incorrect_segment['views'])
|
| 378 |
+
safe_print('\t\tLocked:',
|
| 379 |
+
incorrect_segment['locked'])
|
| 380 |
+
|
| 381 |
+
safe_print('\t\tCurrent Category:',
|
| 382 |
+
incorrect_segment['category'])
|
| 383 |
+
safe_print('\t\tPredicted Category:',
|
| 384 |
+
incorrect_segment['predicted'])
|
| 385 |
+
safe_print('\t\tProbabilities:')
|
| 386 |
for item in incorrect_segment['scores']:
|
| 387 |
+
safe_print(
|
| 388 |
f"\t\t\t{item['label']}: {item['score']}")
|
| 389 |
|
| 390 |
+
safe_print()
|
| 391 |
|
| 392 |
except KeyboardInterrupt:
|
| 393 |
pass
|