kernel_id int64 24.2k 23.3M | prompt stringlengths 8 1.85M | completetion stringlengths 1 182k | comp_name stringlengths 5 57 |
|---|---|---|---|
14,958,171 | SEED=42
seed_everything(SEED)
APEX = False
ACCUM = 2
SWAP_VALID_AND_TRAIN = False
N_VIZ = 10
USE_NMS = False
SCORE_THRESHOLD = 0.65
NMS_IOU_THRESHOLD = 0.5
IMG_SIZE = 1024
WBF_IOU, WBF_SKIP_BOX = 0.44, 0.38
PP_SHRINK = [-1,0]
WBF_SCORE_THRESHOLD = 0.265
USE_BOUNDS_FILTER = True
LOWER_BOUND, UPPER_BOUND = 70, 175000
... | model.eval()
predictions = []
for batch in test_dataloader:
batch = tuple(t.to(device)for t in batch)
b_input_ids, b_mask = batch
with torch.no_grad() :
logits = model(b_input_ids, token_type_ids=None, attention_mask = b_mask)
logits = logits.detach().cpu().numpy()
predictions.append(logits)
test_predictions = [item... | Natural Language Processing with Disaster Tweets |
14,958,171 | <import_modules><EOS> | submission = pd.DataFrame({'id':test_df['id'], 'target':test_predictions})
submission.to_csv('./submission.csv', index=False ) | Natural Language Processing with Disaster Tweets |
14,910,610 | <SOS> metric: meanfscore Kaggle data source: natural-language-processing-with-disaster-tweets<set_options> | import numpy as np
import pandas as pd
import tensorflow as tf
import csv
import re | Natural Language Processing with Disaster Tweets |
14,910,610 | %%writefile./modeling/wheat_detector.py
class WheatDetector(nn.Module):
def __init__(self, cfg, **kwargs):
super(WheatDetector, self ).__init__()
self.backbone = resnest_fpn_backbone(pretrained=False)
self.base = FasterRCNN(self.backbone, num_classes=cfg.MODEL.NUM_CLASSES, **kwargs)
def forward(self, images, targets=... | def convert_test_attributes_to_onehot(df, value_list):
one_hots = np.zeros(( len(df), len(value_list)))
for index, row in df.iterrows() :
if row['keyword'] != '':
try:
one_hots[index, value_list.index(row['keyword'])] = 1
except ValueError:
continue
if row['keyword'] != '':
try:
one_hots[index, value_list.index(row['l... | Natural Language Processing with Disaster Tweets |
14,910,610 | sys.path.insert(0, "./external/wbf")
warnings.filterwarnings("ignore")
class BaseWheatTTA:
image_size = IMG_SIZE
def augment(self, image):
raise NotImplementedError
def batch_augment(self, images):
raise NotImplementedError
def deaugment_boxes(self, boxes):
raise NotImplementedError
class TTAReduceSaturation(BaseWh... | def build_model(lstm_shape, dense_shape):
dropout = 0.8
dense_input = tf.keras.layers.Input(shape=(dense_shape))
dense1 = tf.keras.layers.Dense(50 )(dense_input)
lstm_input = tf.keras.layers.Input(shape=(lstm_shape))
lstm1 = tf.keras.layers.GaussianNoise(0.075 )(lstm_input)
lstm2 = tf.keras.layers.LSTM(units=500, ret... | Natural Language Processing with Disaster Tweets |
14,910,610 | class Tester:
def __init__(self, models, device, cfg, test_loader, n_viz=N_VIZ):
self.config = cfg
self.test_loader = test_loader
self.base_dir = f'{self.config.OUTPUT_DIR}'
if not os.path.exists(self.base_dir):
os.makedirs(self.base_dir)
self.log_path = f'{self.base_dir}/log.txt'
self.score_threshold = SCORE_THRESHOL... | MAX_TWEET_LENGTH = 280
VECTORS_PER_WORD = 50
BATCH_SIZE = 256
NUM_EPOCHS = 40 | Natural Language Processing with Disaster Tweets |
14,910,610 | cfg['OUTPUT_DIR'] = "/kaggle/working/"
cfg['DATASETS']['ROOT_DIR'] = "/kaggle/input/global-wheat-detection"
cfg['TEST']['IMS_PER_BATCH'] = 1
cfg['TEST']['WEIGHT'] = BEST_PATHS
cfg<load_from_csv> | print('Loading test and training data')
test_data = pd.read_csv('/kaggle/input/nlp-getting-started/test.csv')
train_data = pd.read_csv('/kaggle/input/nlp-getting-started/train.csv')
print('Loading GloVe file')
glove_data = pd.read_csv('/kaggle/input/glove50d/glove.6B.50d.txt', sep=' ', index_col=0, header = None, q... | Natural Language Processing with Disaster Tweets |
14,910,610 | <categorify><EOS> | model = build_model([MAX_TWEET_LENGTH, VECTORS_PER_WORD], len(df_one_hot_attributes.columns))
model.summary()
checkpoint_save = tf.keras.callbacks.ModelCheckpoint('saved_model.h5', save_best_only=True, monitor='val_acc', mode='min')
model.fit(x=[df_one_hot_attributes, vectorized_training_data], y=train_data['target'],... | Natural Language Processing with Disaster Tweets |
13,338,682 | <SOS> metric: meanfscore Kaggle data source: natural-language-processing-with-disaster-tweets<split> | import tensorflow as tf
from transformers import BertTokenizer, TFBertModel, AdamWeightDecay
from tensorflow import keras
import matplotlib.pyplot as plt
import string
import re
import numpy as np
import pandas as pd
import emoji
import os
from sklearn.metrics import accuracy_score
from skopt.utils import use_named_arg... | Natural Language Processing with Disaster Tweets |
13,338,682 | marking_list, train_ids_list, valid_ids_list = [], [], []
for ii in range(len(BEST_PATHS)) :
print('
** weights --
cfg.DATASETS.VALID_FOLD = ii
marking, train_ids0, valid_ids0 = split_dataset(cfg)
if SWAP_VALID_AND_TRAIN:
print('swap!!')
valid_ids, train_ids =train_ids0, valid_ids0
else:
train_ids, valid_ids =train_i... | train = pd.read_csv('.. /input/nlp-getting-started/train.csv', header = 0, encoding="utf8")
test = pd.read_csv('.. /input/nlp-getting-started/test.csv', header = 0, encoding="utf8")
train = train.drop(columns=['id', 'keyword', 'location'])
test = test.drop(columns=['id', 'keyword', 'location'])
train = train.drop_d... | Natural Language Processing with Disaster Tweets |
13,338,682 |
<set_options> | def remove_tweet_object(tweet):
tweet = re.sub(r"https?://\S+|www\.\S+", " ", tweet)
tweet = re.sub(r"
tweet = re.sub(r"@\w+", " ", tweet)
tweet = emoji.get_emoji_regexp().sub(" ", tweet)
return tweet
def text_filter(tweet):
tweet = tweet.lower()
tweet = re.sub(r"’", "'", tweet)
tweet = remove_tweet_object(tweet)
... | Natural Language Processing with Disaster Tweets |
13,338,682 | %%writefile./data/transforms/build.py
def get_train_transforms(cfg):
return A.Compose(
[
A.Resize(1024, 1024, p=1.0),
A.RandomSizedCrop(min_max_height=cfg.INPUT.RSC_MIN_MAX_HEIGHT, height=cfg.INPUT.RSC_HEIGHT,
width=cfg.INPUT.RSC_WIDTH, p=cfg.INPUT.RSC_PROB),
A.OneOf([
A.HueSaturationValue(hue_shift_limit=cfg.INPUT.HS... | X_train, X_val, y_train, y_val = train_test_split(train.text, train.target, test_size=0.2, random_state=42 ) | Natural Language Processing with Disaster Tweets |
13,338,682 | warnings.filterwarnings("ignore")
def build_dataset(cfg, marking,train_ids, valid_ids):
train_dataset = train_wheat(
root = cfg.DATASETS.ROOT_DIR,
image_ids=train_ids,
marking=marking,
transforms=build_transforms(cfg, is_train=True),
test=False,
)
validation_dataset = train_wheat(
root=cfg.DATASETS.ROOT_DIR,
image... | name_bert = "bert-base-uncased"
tokenizer = BertTokenizer.from_pretrained(name_bert, do_lower_case=True)
bert_model = TFBertModel.from_pretrained(name_bert, output_hidden_states=True, trainable=True)
bert_w = bert_model.get_weights() | Natural Language Processing with Disaster Tweets |
13,338,682 | class Fitter:
def __init__(self, model, device, cfg, train_loader, val_loader, logger, mixed_precision=APEX, accum=ACCUM):
self.config = cfg
self.epoch = 0
self.train_loader = train_loader
self.val_loader = val_loader
self.base_dir = f'{self.config.OUTPUT_DIR}'
if not os.path.exists(self.base_dir):
os.makedirs(self.bas... | combined = pd.concat([X_train, X_val, test.text], axis=0)
combined = tokenizer(combined.values.tolist() , padding=True, truncation=True, return_tensors='tf')
train_input =(combined["input_ids"][:len(X_train)], combined["attention_mask"][:len(X_train)], combined["token_type_ids"][:len(X_train)])
val_input =(combined[... | Natural Language Processing with Disaster Tweets |
13,338,682 | cfg.defrost()
cfg['DATASETS']['ROOT_DIR'] = NEW_INPUT_PATH
cfg.INPUT.HSV_H = HSV_H
cfg.INPUT.HSV_S = HSV_S
cfg.INPUT.HSV_V = HSV_V
cfg.INPUT.BC_B = BC_B
cfg.INPUT.BC_C = BC_C
cfg.INPUT.COTOUT_NUM_HOLES=0
cfg.SOLVER.BASE_LR = BASE_LR
cfg.SOLVER.BIAS_LR_FACTOR = BIAS_LR_FACTOR
cfg.SOLVER.MOMENTUM=MOMENTUM
cfg.SOLVER.WARM... | bert_model.set_weights(bert_w)
optimizer = AdamWeightDecay(learning_rate=3e-5, weight_decay_rate=0.01, exclude_from_weight_decay=["LayerNorm", "layer_norm", "bias"])
loss = keras.losses.BinaryCrossentropy(from_logits=False, label_smoothing=0.0)
input_word_ids = keras.layers.Input(shape=(train_input[0].shape[1],), dt... | Natural Language Processing with Disaster Tweets |
13,338,682 | fitters=[]
for ii,path in enumerate(cfg['TEST']['WEIGHT']):
cfg['OUTPUT_DIR'] = OUTPUT_DIRS[ii]
checkpoint = torch.load(path)
cfg.SOLVER.MAX_EPOCHS = checkpoint['epoch']+PSEUDO_EPOCHS+1
if n_test <11:
cfg.SOLVER.MAX_EPOCHS = checkpoint['epoch']+PSEUDO_EPOCHS_COMMIT+1
print('epochs = %d+%d+%d'%(checkpoint['epoch'],PSEU... | model_bert_enc = keras.models.Model(inputs=[input_word_ids, input_mask, segment_ids], outputs=relu)
train_text_vect = model_bert_enc.predict(train_input)
val_text_vect = model_bert_enc.predict(val_input)
test_text_vect = model_bert_enc.predict(test_input)
print("Done" ) | Natural Language Processing with Disaster Tweets |
13,338,682 | best_paths = []
for ii in range(len(OUTPUT_DIRS)) :
if os.path.exists(OUTPUT_DIRS[ii]+'best-checkpoint.bin'):
best_path = OUTPUT_DIRS[ii]+'best-checkpoint.bin'
elif os.path.exists(OUTPUT_DIRS[ii]+'last-checkpoint.bin'):
best_path = OUTPUT_DIRS[ii]+'last-checkpoint.bin'
else:
best_path = BEST_PATHS[ii]
best_paths.append... | space = [Real(1, 1e3, prior="log-uniform", name="C_svm", transform="identity"),
Real(1, 1e3, prior="log-uniform", name="C_lr", transform="identity"),
Categorical(['hinge', 'squared_hinge'], name="loss_svm", transform="identity"),
Categorical(['l1', 'l2', 'elasticnet'], name="penalty_sgd", transform="identity"),
Categor... | Natural Language Processing with Disaster Tweets |
13,338,682 | if True:
cfg['OUTPUT_DIR'] = "/kaggle/working/"
cfg['DATASETS']['ROOT_DIR'] = "/kaggle/input/global-wheat-detection"
cfg['TEST']['WEIGHT'] = best_paths
cfg['TEST']['IMS_PER_BATCH'] = 1
print(cfg)
test_loader = make_test_data_loader(cfg)
tester = Tester(models=models, device=device, cfg=cfg, test_loader=test_loader, n... | estimators = [
('1', make_pipeline(MinMaxScaler() , LinearSVC(C=res.x[0], loss=res.x[2] ,random_state=42))),
('2', make_pipeline(MinMaxScaler() , GaussianNB())) ,
('3', make_pipeline(MinMaxScaler() , SGDClassifier(penalty=res.x[3], loss=res.x[4],random_state=42))),
]
clf = StackingClassifier(estimators=estimators, f... | Natural Language Processing with Disaster Tweets |
13,338,682 | <set_options><EOS> | df_submission = pd.read_csv('.. /input/nlp-getting-started/sample_submission.csv', index_col=0 ).fillna('')
df_submission['target'] = clf.predict(test_text_vect)
df_submission.to_csv('submission.csv')
df_submission | Natural Language Processing with Disaster Tweets |
14,011,004 | <SOS> metric: meanfscore Kaggle data source: natural-language-processing-with-disaster-tweets<load_from_csv> | for dirname, _, filenames in os.walk('/kaggle/input'):
for filename in filenames:
print(os.path.join(dirname, filename))
| Natural Language Processing with Disaster Tweets |
14,011,004 | def load_dataset(root):
csv = pd.read_csv(os.path.join(root, "train.csv"))
data = {}
for i in csv.index:
key = csv["image_id"][i]
bbox = json.loads(csv["bbox"][i])
bbox = [bbox[0], bbox[1], bbox[0] + bbox[2], bbox[1] + bbox[3], 0.0]
if key in data:
data[key].append(bbox)
else:
data[key] = [bbox]
return sorted(
[(k, ... | tweets = pd.read_csv('/kaggle/input/nlp-getting-started/train.csv')
test = pd.read_csv('/kaggle/input/nlp-getting-started/test.csv' ) | Natural Language Processing with Disaster Tweets |
14,011,004 | def load_model(path, ctx=mx.cpu()):
net = gcv.model_zoo.yolo3_darknet53_custom(["wheat"], pretrained_base=False)
net.set_nms(post_nms=150)
net.load_parameters(path, ctx=ctx)
return net
<normalization> | stop = set(stopwords.words('english'))
corpus0 = []
[corpus0.append(word.lower())for tweet in tweets[tweets.target == 0].text for word in word_tokenize(tweet)]
corpus1 = []
[corpus1.append(word.lower())for tweet in tweets[tweets.target == 1].text for word in word_tokenize(tweet)]
def count_top_stopwords(corpus):
stopwo... | Natural Language Processing with Disaster Tweets |
14,011,004 | def inference(models, path):
raw = load_image(path)
rh, rw, _ = raw.shape
classes_list = []
scores_list = []
bboxes_list = []
for _ in range(16):
img, flips = gcv.data.transforms.image.random_flip(raw, px=0.5, py=0.5)
x, _ = gcv.data.transforms.presets.yolo.transform_test(img, short=img_s)
_, _, xh, xw = x.shape
rot... | stop = ENGLISH_STOP_WORDS.union(stop)
def remove_url(txt):
return ' '.join(re.sub('([^0-9A-Za-z \t])|(\w+:\/\/\S+)', '', txt ).split())
corpus0 = []
[corpus0.append(word.lower())for tweet in tweets[tweets.target == 0].text for word in word_tokenize(remove_url(tweet)) ]
corpus0 = list(filter(lambda x: x not in stop, c... | Natural Language Processing with Disaster Tweets |
14,011,004 | rounds = 2
max_epochs = 5
learning_rate = 0.001
batch_size = 2
img_s = 1024
threshold = 0.1
context = mx.gpu()
print("Loading pre-trained model...")
model = load_model(".. /input/yolov3/global-wheat-yolo3-darknet53_best2.params", ctx=context)
print("Loading training set...")
dataset = load_dataset("/kaggle/input/glo... | tweets['polarity'] = [TextBlob(tweet ).sentiment.polarity for tweet in tweets.text]
tweets['subjectivity'] = [TextBlob(tweet ).sentiment.subjectivity for tweet in tweets.text]
tweets['exclaimation_num'] = [tweet.count('!')for tweet in tweets.text]
tweets['questionmark_num'] = [tweet.count('?')for tweet in tweets.text]
... | Natural Language Processing with Disaster Tweets |
14,011,004 | sys.path.insert(0, "/kaggle/input/weightedboxesfusion")
<define_variables> | tweets.keyword.fillna('None', inplace=True)
def decontraction(phrase):
phrase = re.sub(r"won't", "will not", phrase)
phrase = re.sub(r"can't", "can not", phrase)
phrase = re.sub(r"n't", " not", phrase)
phrase = re.sub(r"'re", " are", phrase)
phrase = re.sub(r"'s", " is", phrase)
phrase = re.sub(r"'d", " would", p... | Natural Language Processing with Disaster Tweets |
14,011,004 | DATA_DIR = "/kaggle/input/global-wheat-detection"
MODELS_IN_DIR = "/kaggle/input/frcnn152foldthree"<load_from_csv> | tweets.text = tweets.text.apply(lambda x: remove_url(x))
def remove_punct(text):
new_punct = re.sub('\ |\!|\?', '', punctuation)
table = str.maketrans('','', new_punct)
return text.translate(table)
tweets.text = tweets.text.apply(lambda x: remove_punct(x))
def replace_amp(text):
text = re.sub(r' amp ', ' and ', text... | Natural Language Processing with Disaster Tweets |
14,011,004 | test_df = pd.read_csv(os.path.join(DATA_DIR, "sample_submission.csv"))
test_df.shape<feature_engineering> | lemmatizer = WordNetLemmatizer()
def lemma(text):
words = word_tokenize(text)
return ' '.join([lemmatizer.lemmatize(w.lower() , pos='v')for w in words])
tweets.text = tweets.text.apply(lambda x: lemma(x)) | Natural Language Processing with Disaster Tweets |
14,011,004 | class WheatDataset(Dataset):
def __init__(self, dataframe, image_dir, transforms=None):
super().__init__()
self.image_ids = dataframe['image_id'].unique()
self.df = dataframe
self.image_dir = image_dir
self.transforms = transforms
def __len__(self)-> int:
return len(self.image_ids)
def __getitem__(self, idx: int):
ima... | def generate_ngrams(text, n):
words = word_tokenize(text)
return [' '.join(ngram)for ngram in list(get_data(ngrams(words, n)))if not all(w in stop for w in ngram)]
def get_data(gen):
try:
for elem in gen:
yield elem
except(RuntimeError, StopIteration):
return | Natural Language Processing with Disaster Tweets |
14,011,004 | def get_model() :
backbone = resnet_fpn_backbone('resnet152', pretrained=False)
model = FasterRCNN(backbone, num_classes=2)
return model<load_pretrained> | bigrams_disaster = tweets[tweets.target==1].text.apply(lambda x: generate_ngrams(x, 2))
bigrams_ndisaster = tweets[tweets.target==0].text.apply(lambda x: generate_ngrams(x, 2))
bigrams_d_dict = {}
for bgs in bigrams_disaster:
for bg in bgs:
if bg in bigrams_d_dict:
bigrams_d_dict[bg] += 1
else:
bigrams_d_dict[bg] = 1
b... | Natural Language Processing with Disaster Tweets |
14,011,004 | DEVICE = torch.device('cuda')if torch.cuda.is_available() else torch.device('cpu')
model = get_model()
model.load_state_dict(torch.load(os.path.join(MODELS_IN_DIR, "best_model.pth")))
model.eval()
model.to(DEVICE)
1 == 1<categorify> | trigrams_disaster = tweets[tweets.target==1].text.apply(lambda x: generate_ngrams(x, 3))
trigrams_ndisaster = tweets[tweets.target==0].text.apply(lambda x: generate_ngrams(x, 3))
trigrams_d_dict = {}
for tgs in trigrams_disaster:
for tg in tgs:
if tg in trigrams_d_dict:
trigrams_d_dict[tg] += 1
else:
trigrams_d_dict[tg... | Natural Language Processing with Disaster Tweets |
14,011,004 | def get_test_transforms() :
return A.Compose([
A.Resize(height=1024, width=1024, p=1.0),
ToTensorV2(p=1.0),
], p=1.0 )<create_dataframe> | def remove_stopwords(text):
word_tokens = word_tokenize(text)
return ' '.join([w.lower() for w in word_tokens if not w.lower() in stop])
tweets['text_nostopwords'] = tweets.text.apply(lambda x: remove_stopwords(x)) | Natural Language Processing with Disaster Tweets |
14,011,004 | def collate_fn(batch):
return tuple(zip(*batch))
test_dataset = WheatDataset(test_df, os.path.join(DATA_DIR, "test"), get_test_transforms())
test_data_loader = DataLoader(
test_dataset,
batch_size=4,
shuffle=False,
num_workers=1,
drop_last=False,
collate_fn=collate_fn
)<define_variables> | mask = np.array(Image.open('.. /input/twitterlogo3/twitter-logo-png-transparent.png'))
reverse = mask[...,::-1,:]
def wc_words(target, mask=mask):
words = [word.lower() for tweet in tweets[tweets.target == target].text_nostopwords for word in tweet.split() ]
words = list(filter(lambda w: w != 'like', words))
words = li... | Natural Language Processing with Disaster Tweets |
14,011,004 | def format_prediction_string(boxes, scores):
pred_strings = []
for j in zip(scores, boxes):
pred_strings.append("{0:.4f} {1} {2} {3} {4}".format(j[0], j[1][0], j[1][1], j[1][2], j[1][3]))
return " ".join(pred_strings )<categorify> | pd.reset_option('max_colwidth')
tweets.drop('text_nostopwords', axis=1, inplace=True)
tweets.head() | Natural Language Processing with Disaster Tweets |
14,011,004 | class BaseWheatTTA:
image_size = 1024
def augment(self, image):
raise NotImplementedError
def batch_augment(self, images):
raise NotImplementedError
def deaugment_boxes(self, boxes):
raise NotImplementedError
class TTAHorizontalFlip(BaseWheatTTA):
def augment(self, image):
return image.flip(1)
def batch_augment(se... | X_train, X_val, y_train, y_val = train_test_split(tweets.drop(['keyword','location','target'],axis=1), tweets[['target']], test_size=0.2, stratify=tweets[['target']], random_state=0)
X_train_text = X_train['text']
X_val_text = X_val['text']
print('X_train shape: ', X_train.shape)
print('X_val shape: ', X_val.shape)
... | Natural Language Processing with Disaster Tweets |
14,011,004 | tta_transforms = []
for tta_combination in product([TTAHorizontalFlip() , None],
[TTAVerticalFlip() , None],
[TTARotate90() , None]):
tta_transforms.append(TTACompose([tta_transform for tta_transform in tta_combination if tta_transform]))<categorify> | print('Train Class Proportion:
', y_train['target'].value_counts() / len(y_train)* 100)
print('
Validation Class Proportion:
', y_val['target'].value_counts() / len(y_val)* 100 ) | Natural Language Processing with Disaster Tweets |
14,011,004 | def make_tta_predictions(images, score_threshold=0.5):
with torch.no_grad() :
images = torch.stack(images ).float().to(DEVICE)
predictions = []
for tta_transform in tta_transforms:
result = []
outputs = model(tta_transform.batch_augment(images.clone()))
for i, image in enumerate(images):
boxes = outputs[i]['boxes'].da... | tokenizer_1 = Tokenizer(num_words=5000, oov_token='<UNK>')
tokenizer_1.fit_on_texts(X_train_text ) | Natural Language Processing with Disaster Tweets |
14,011,004 | def run_wbf(predictions, image_index, image_size=1024, iou_thr=0.5, skip_box_thr=0.43, weights=None):
boxes = [(prediction[image_index]['boxes']/(image_size-1)).tolist() for prediction in predictions]
scores = [prediction[image_index]['scores'].tolist() for prediction in predictions]
labels = [np.ones(prediction[image_... | X_train_text = tokenizer_1.texts_to_sequences(X_train_text)
X_val_text = tokenizer_1.texts_to_sequences(X_val_text)
print(X_train_text[:10])
print('')
print(X_val_text[:10] ) | Natural Language Processing with Disaster Tweets |
14,011,004 | results = []
for images, image_ids in test_data_loader:
predictions = make_tta_predictions(images)
for i, image in enumerate(images):
boxes, scores, labels = run_wbf(predictions, image_index=i)
boxes = boxes.round().astype(np.int32 ).clip(min=0, max=1023)
image_id = image_ids[i]
boxes[:, 2] = boxes[:, 2] - boxes[:, ... | tokenizer_1.sequences_to_texts([X_train_text[1]] ) | Natural Language Processing with Disaster Tweets |
14,011,004 | test_df = pd.DataFrame(results, columns=['image_id', 'PredictionString'])
test_df.head()<save_to_csv> | vocab_size = len(tokenizer_1.word_index)+ 1
embeddings_index = dict()
f = open('.. /input/glovetwitter27b100dtxt/glove.twitter.27B.200d.txt')
for line in f:
values = line.split()
word = values[0]
coefs = np.asarray(values[1:], dtype='float32')
embeddings_index[word] = coefs
f.close()
print('Loaded %s word vectors.' %... | Natural Language Processing with Disaster Tweets |
14,011,004 | test_df.to_csv('submission.csv', index=False )<install_modules> | embedding_matrix = np.zeros(( vocab_size, 200))
for word, i in tokenizer_1.word_index.items() :
embedding_vector = embeddings_index.get(word)
if embedding_vector is not None:
embedding_matrix[i] = embedding_vector
print('Embedding Matrix Shape:', embedding_matrix.shape ) | Natural Language Processing with Disaster Tweets |
14,011,004 | !pip install --no-deps '.. /input/timm0130/timm-0.1.30-py3-none-any.whl' > /dev/null<install_modules> | num_epochs=15
dropout=0.2
recurrent_dropout=0.2
lr=0.0005
batch_size=128
class_weight = {0: y_train['target'].value_counts() [1]/len(y_train), 1: y_train['target'].value_counts() [0]/len(y_train)} | Natural Language Processing with Disaster Tweets |
14,011,004 | !pip install --no-deps '.. /input/pycocotools/pycocotools-2.0-cp37-cp37m-linux_x86_64.whl' > /dev/null<set_options> | lstm_model = Sequential()
embedding_layer = Embedding(vocab_size, 200, weights=[embedding_matrix], input_length=maxlen, trainable=False)
lstm_model.add(embedding_layer)
lstm_model.add(LSTM(128, return_sequences=True, dropout=dropout, recurrent_dropout=recurrent_dropout))
lstm_model.add(LSTM(128))
lstm_model.add(Dense... | Natural Language Processing with Disaster Tweets |
14,011,004 | %matplotlib inline
VerticalFlip, HorizontalFlip, IAASharpen,
OneOf, Compose , BboxParams, Resize, HueSaturationValue
,RandomBrightnessContrast, ToGray , Cutout , RandomSizedCrop )<import_modules> | checkpoint = ModelCheckpoint('lstm_model.h5', monitor='val_acc', save_best_only=True)
history = lstm_model.fit(X_train_text, y_train, batch_size=batch_size, callbacks=[checkpoint], epochs=num_epochs,
class_weight=class_weight, validation_data=(X_val_text, y_val), verbose=1)
plot_model_performance(history ) | Natural Language Processing with Disaster Tweets |
14,011,004 | from effdet import get_efficientdet_config, EfficientDet, DetBenchTrain , DetBenchPredict
from effdet.efficientdet import HeadNet<define_variables> | test['char_len'] = test.text.str.len()
word_tokens = [len(word_tokenize(tweet)) for tweet in test.text]
test['word_len'] = word_tokens
sent_tokens = [len(sent_tokenize(tweet)) for tweet in test.text]
test['sent_len'] = sent_tokens | Natural Language Processing with Disaster Tweets |
14,011,004 | DIR_PATH = '/kaggle/input/global-wheat-detection/'
dir = glob.glob(os.path.join(DIR_PATH , '*'))
dir.sort(reverse=True)
train_paths = glob.glob(os.path.join(dir[1] , '*'))
test_paths = glob.glob(os.path.join(dir[2] , '*'))<feature_engineering> | test['polarity'] = [TextBlob(tweet ).sentiment.polarity for tweet in test.text]
test['subjectivity'] = [TextBlob(tweet ).sentiment.subjectivity for tweet in test.text]
test['exclaimation_num'] = [tweet.count('!')for tweet in test.text]
test['questionmark_num'] = [tweet.count('?')for tweet in test.text]
def count_url_ha... | Natural Language Processing with Disaster Tweets |
14,011,004 | df = pd.read_csv(dir[0])
bboxs = np.stack(df['bbox'].apply(lambda x: np.fromstring(x[1:-1], sep=',')))
for i, column in enumerate(['x', 'y', 'w', 'h']):
df[column] = bboxs[:,i]
df.drop(columns=['bbox'], inplace=True)
df<feature_engineering> | test.keyword.fillna('None', inplace=True)
def decontraction(phrase):
phrase = re.sub(r"won't", "will not", phrase)
phrase = re.sub(r"can't", "can not", phrase)
phrase = re.sub(r"n't", " not", phrase)
phrase = re.sub(r"'re", " are", phrase)
phrase = re.sub(r"'s", " is", phrase)
phrase = re.sub(r"'d", " would", phr... | Natural Language Processing with Disaster Tweets |
14,011,004 | skf = StratifiedKFold(n_splits=5, shuffle=True, random_state=42)
df_folds = df[['image_id']].copy()
df_folds.loc[:, 'bbox_count'] = 1
df_folds = df_folds.groupby('image_id' ).count()
df_folds.loc[:, 'source'] = df[['image_id', 'source']].groupby('image_id' ).min() ['source']
df_folds.loc[:, 'stratify_group'] = np.char... | test.text = test.text.apply(lambda x: remove_url(x))
def remove_punct(text):
new_punct = re.sub('\ |\!|\?', '', punctuation)
table=str.maketrans('','',new_punct)
return text.translate(table)
test.text = test.text.apply(lambda x: remove_punct(x))
def replace_amp(text):
text = re.sub(r" amp ", " and ", text)
return t... | Natural Language Processing with Disaster Tweets |
14,011,004 | image_id = '8425a537b.jpg'
image_path = glob.glob(os.path.join(dir[1] , image_id))
image , boxes = load_image_and_boxes(image_path[0])
show_image(image, boxes, "Image without bounding box" )<train_model> | lemmatizer = WordNetLemmatizer()
def lemma(text):
words = word_tokenize(text)
return ' '.join([lemmatizer.lemmatize(w.lower() , pos='v')for w in words])
test.text = test.text.apply(lambda x: lemma(x)) | Natural Language Processing with Disaster Tweets |
14,011,004 | image_id = 'b3c96d5ad.jpg'
image_path = glob.glob(os.path.join(dir[1] , image_id))
image , boxes = load_image_and_boxes(image_path[0])
show_image(image, boxes, "Image with bounding box" )<normalization> | test_text = test['text']
test_text = tokenizer_1.texts_to_sequences(test_text)
test_text = pad_sequences(test_text, padding='post', maxlen=50)
print('X_test shape:', test_text.shape ) | Natural Language Processing with Disaster Tweets |
14,011,004 | <normalization><EOS> | lstm_model.load_weights('lstm_model.h5')
submission = test.copy() [['id']]
submission['target'] = lstm_model.predict_classes(test_text)
submission.to_csv('submission.csv', index=False)
display(submission.head() ) | Natural Language Processing with Disaster Tweets |
21,936,667 | <SOS> metric: meanfscore Kaggle data source: natural-language-processing-with-disaster-tweets<feature_engineering> | train = pd.read_csv(".. /input/nlp-getting-started/train.csv")
test = pd.read_csv(".. /input/nlp-getting-started/test.csv")
train.head() | Natural Language Processing with Disaster Tweets |
21,936,667 | class WheatDataset(Dataset):
def __init__(self , dataframe , image_ids, transforms = None):
super().__init__()
self.image_ids = image_ids
self.dataframe = dataframe
self.transforms = transforms
def __getitem__(self, index: int):
image_id = self.image_ids[index]
image, boxes = self.load_image_and_boxes(index)
labels = ... | !wget --quiet https://raw.githubusercontent.com/tensorflow/models/master/official/nlp/bert/tokenization.py
!pip install sentencepiece | Natural Language Processing with Disaster Tweets |
21,936,667 | def collate_fn(batch):
return tuple(zip(*batch))<create_dataframe> | import tensorflow_hub as hub
import tokenization
from sklearn.model_selection import train_test_split
import tensorflow as tf | Natural Language Processing with Disaster Tweets |
21,936,667 | bs = 2
fold_number = 0
train_set = WheatDataset(dataframe=df, image_ids=df_folds[df_folds['fold'] != fold_number].index.values , transforms=transforms_train)
valid_set = WheatDataset(dataframe=df, image_ids=df_folds[df_folds['fold'] == fold_number].index.values , transforms=transforms_valid)
train_loader = DataLoader... | X = np.array(train['text'])
y = np.array(train['target'] ) | Natural Language Processing with Disaster Tweets |
21,936,667 | class Training:
def __init__(self, model, device, config):
self.config = config
self.epoch = 0
self.base_dir = f'{config.folder}'
if not os.path.exists(self.base_dir):
os.makedirs(self.base_dir)
self.best_calc_loss = 10**5
self.model = model
self.device = device
self.optimizer = torch.optim.AdamW(self.model.parameters... | def bert_encode(texts,tokenizer, max_len = 512):
all_tokens = []
all_masks = []
all_segments = []
for text in texts:
text = tokenizer.tokenize(text)
text = text[:max_len-2]
input_sequence = ["[CLS]"] + text + ["[SEP]"]
tokens = tokenizer.convert_tokens_to_ids(input_sequence)
pad_len = max_len - len(input_sequence)
t... | Natural Language Processing with Disaster Tweets |
21,936,667 | def get_model(num_classes = 1):
config = get_efficientdet_config('tf_efficientdet_d7x')
model = EfficientDet(config, pretrained_backbone=False)
checkpoint = torch.load('.. /input/efficientdetd7x/tf_efficientdet_d7x-f390b87c.pth')
model.load_state_dict(checkpoint)
config.num_classes = num_classes
config.image_size =... | def build_model(bert_layer,max_len=512):
input_word_ids = tf.keras.layers.Input(shape=(max_len,),dtype=tf.int32,name="input_word_ids")
input_mask = tf.keras.layers.Input(shape=(max_len,),dtype=tf.int32,name="input_mask")
input_segment_ids = tf.keras.layers.Input(shape=(max_len,),dtype=tf.int32,name="input_segment_ids... | Natural Language Processing with Disaster Tweets |
21,936,667 | class GlobalParametersTrain:
lr = 0.0002
n_epochs = 20
folder = '.. /input/modelfasterrcnn'
verbose = True
verbose_step = 10
step_scheduler = False
validation_scheduler = True
SchedulerClass = torch.optim.lr_scheduler.ReduceLROnPlateau
scheduler_params = dict(mode='min',factor=0.5,patience=1,verbose=False, threshold=0.... | %%time
module_url = "https://tfhub.dev/tensorflow/bert_en_uncased_L-24_H-1024_A-16/1"
bert_layer = hub.KerasLayer(module_url, trainable=True ) | Natural Language Processing with Disaster Tweets |
21,936,667 | def load_model(checkpoint_path):
config = get_efficientdet_config('tf_efficientdet_d7x')
model = EfficientDet(config, pretrained_backbone=False)
config.num_classes = 1
config.image_size=512
model.class_net = HeadNet(config, num_outputs=config.num_classes, norm_kwargs=dict(eps=.001, momentum=.01))
checkpoint = torch.l... | vocab_file = bert_layer.resolved_object.vocab_file.asset_path.numpy()
do_lower_case = bert_layer.resolved_object.do_lower_case.numpy()
tokenizer = tokenization.FullTokenizer(vocab_file, do_lower_case ) | Natural Language Processing with Disaster Tweets |
21,936,667 | device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
def make_predictions(images , score_threshold=0.22):
images = torch.stack(images ).cuda().float()
predictions = []
with torch.no_grad() :
outputs = model(images , torch.tensor([1.0] * images.shape[0], dtype=torch.float ).to(device), torch.tensor([im... | X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.2)
train_input = bert_encode(X_train,tokenizer,max_len=264)
val_input = bert_encode(X_test,tokenizer,max_len=264)
test_input = bert_encode(test.text.values,tokenizer,max_len=264)
train_labels = y_train
val_labels = y_test | Natural Language Processing with Disaster Tweets |
21,936,667 | def run_wbf(predictions, image_index, image_size=512, iou_thr=0.44, skip_box_thr=0.43, weights=None):
boxes = [(prediction[image_index]['boxes']/(image_size-1)).tolist() for prediction in predictions]
scores = [prediction[image_index]['scores'].tolist() for prediction in predictions]
labels = [prediction[image_index]['... | checkpoint = ModelCheckpoint('model.h5', monitor='val_loss', save_best_only=True)
es = EarlyStopping(monitor='val_loss',patience=3,verbose=1,restore_best_weights=True,min_delta=0.01)
model.compile(optimizer=Adam(lr=1e-5),loss='binary_crossentropy',metrics=['accuracy'])
train_history = model.fit(
train_input, train_... | Natural Language Processing with Disaster Tweets |
21,936,667 | for j,(images, targets , image_ids)in enumerate(valid_loader):
break
predictions = make_predictions(images)
i = 0
sample = images[i].permute(1,2,0 ).cpu().numpy()
boxes, scores, labels = run_wbf(predictions, image_index=i)
boxes = boxes.astype(np.int32 ).clip(min=0, max=511 )<categorify> | y_pred = model.predict(test_input)
ans = pd.DataFrame({'id':np.array(test['id']),'target':np.array(y_pred.round().astype(int)).reshape(-1)})
ans.to_csv('submission.csv',index=False)
ans | Natural Language Processing with Disaster Tweets |
7,928,811 | class BaseWheatTTA:
image_size = 512
def augment(self, image):
raise NotImplementedError
def batch_augment(self, images):
raise NotImplementedError
def deaugment_boxes(self, boxes):
raise NotImplementedError
class TTAHorizontalFlip(BaseWheatTTA):
def augment(self, image):
return image.flip(1)
def batch_augment(sel... | nltk.download('wordnet')
nltk.download('punkt')
| Natural Language Processing with Disaster Tweets |
7,928,811 | tta_transforms = []
for tta_combination in product([TTAHorizontalFlip() , None], [TTAVerticalFlip() , None],[TTARotate90() , None]):
tta_transforms.append(TTACompose([tta_transform for tta_transform in tta_combination if tta_transform]))<categorify> | df = pd.read_csv("/kaggle/input/nlp-getting-started/train.csv")
df_test = pd.read_csv("/kaggle/input/nlp-getting-started/test.csv")
df.sample(10 ) | Natural Language Processing with Disaster Tweets |
7,928,811 | def make_tta_predictions(images, score_threshold=0.5):
with torch.no_grad() :
images = torch.stack(images ).float().cuda()
predictions = []
for tta_transform in tta_transforms:
result = []
outputs = model(tta_transform.batch_augment(images.clone()), torch.tensor([1]*images.shape[0] ).float().cuda() , torch.tensor([imag... | print("Training :")
print("Length of the data :", len(df))
print(df.isnull().sum() ) | Natural Language Processing with Disaster Tweets |
7,928,811 | test_transforms = Compose([
Resize(height=512, width=512, p=1.0),
ToTensorV2(p=1.0),
], p=1.0 )<categorify> | print("Test :")
print("Length of the data :", len(df_test))
print(df_test.isnull().sum() ) | Natural Language Processing with Disaster Tweets |
7,928,811 | class TestDataset(Dataset):
def __init__(self, image_ids, transforms=None):
super().__init__()
self.image_ids = image_ids
self.transforms = transforms
def __getitem__(self, index):
image_id = self.image_ids[index]
image = cv2.imread(f'{dir[2]}/{image_id}.jpg', cv2.IMREAD_COLOR)
image = cv2.cvtColor(image, cv2.COLOR_BG... | tokens = word_tokenize(df["text"][0])
tokens = [word.lower() for word in tokens]
print(df["text"][0])
print(tokens ) | Natural Language Processing with Disaster Tweets |
7,928,811 | def collate_fn(batch):
return tuple(zip(*batch))<create_dataframe> | words = [word for word in tokens if word.isalpha() ]
print(words ) | Natural Language Processing with Disaster Tweets |
7,928,811 | test_set = TestDataset(image_ids=np.array([path.split('/')[-1][:-4] for path in test_paths]),transforms=test_transforms)
test_loader = DataLoader(test_set,batch_size=4,shuffle=False,num_workers=2,drop_last=False,collate_fn=collate_fn )<define_variables> | stop_words = set(stopwords.words("english"))
words = [word for word in words if not word in stop_words]
print(words ) | Natural Language Processing with Disaster Tweets |
7,928,811 | def format_prediction_string(boxes, scores):
pred_strings = []
for j in zip(scores, boxes):
pred_strings.append("{0:.4f} {1} {2} {3} {4}".format(j[0], j[1][0], j[1][1], j[1][2], j[1][3]))
return " ".join(pred_strings )<save_to_csv> | porter = PorterStemmer()
stemmed = [porter.stem(word)for word in words]
print(stemmed ) | Natural Language Processing with Disaster Tweets |
7,928,811 | SUBMISSION_PATH = '/kaggle/working'
submission_id = 'submission'
submission_path = os.path.join(SUBMISSION_PATH, '{}.csv'.format(submission_id))
sample_submission = pd.DataFrame(submission, columns=["image_id","PredictionString"])
sample_submission.to_csv(submission_path, index=False)
submission_df = pd.read_csv(subm... | lemmatizer = WordNetLemmatizer()
lemmatized = [lemmatizer.lemmatize(word)for word in words]
print(lemmatized ) | Natural Language Processing with Disaster Tweets |
7,928,811 | all_path = glob('.. /input/global-wheat-detection/test/*')
DATA_ROOT_PATH = '.. /input/global-wheat-detection/test'<categorify> | def remove_URL(text):
url = re.compile(r'https?://\S+|www\.\S+')
return url.sub(r'',text)
def remove_html(text):
html=re.compile(r'<.*?>')
return html.sub(r'',text)
def remove_emoji(text):
emoji_pattern = re.compile("["
u"\U0001F600-\U0001F64F"
u"\U0001F300-\U0001F5FF"
u"\U0001F680-\U0001F6FF"
u"\U0001F1E0-\U0001F1... | Natural Language Processing with Disaster Tweets |
7,928,811 | def get_valid_transforms() :
return A.Compose([
A.Resize(height=1024, width=1024, p=1.0),
ToTensorV2(p=1.0),
], p=1.0 )<data_type_conversions> | def join_list(tab):
return " ".join(tab)
df["text_preprocessed"] = df["tokens"].apply(join_list)
df_test["text_preprocessed"] = df_test["tokens"].apply(join_list)
def transform_keyword(word):
return word.split('%20')
df["keyword"] = df.keyword.fillna(" ")
df_test["keyword"] = df_test.keyword.fillna(" ")
df["keywo... | Natural Language Processing with Disaster Tweets |
7,928,811 | class DatasetRetriever(Dataset):
def __init__(self, image_ids, transforms=None):
super().__init__()
self.image_ids = image_ids
self.transforms = transforms
def __getitem__(self, index: int):
image_id = self.image_ids[index]
image = cv2.imread(f'{DATA_ROOT_PATH}/{image_id}.jpg', cv2.IMREAD_COLOR)
image = cv2.cvtColor(i... | X_all = pd.concat([df["text_preprocessed"], df_test["text_preprocessed"]])
sk_doc2bow = CountVectorizer()
sk_doc2bow.fit(X_all)
del X_all
X = sk_doc2bow.transform(df["text_preprocessed"])
X_test = sk_doc2bow.transform(df_test["text_preprocessed"])
X_train, X_val, y_train, y_val = train_test_split(X, df["target"], t... | Natural Language Processing with Disaster Tweets |
7,928,811 | dataset = DatasetRetriever(
image_ids=np.array([path.split('/')[-1][:-4] for path in glob(f'{DATA_ROOT_PATH}/*.jpg')]),
transforms=get_valid_transforms()
)
def collate_fn(batch):
return tuple(zip(*batch))
data_loader = DataLoader(
dataset,
batch_size=4,
shuffle=False,
num_workers=2,
drop_last=False,
collate_fn=coll... | model = SVC(C=2, gamma=0.4, kernel='rbf' ).fit(X_train, y_train)
y_val_pred = model.predict(X_val)
print(accuracy_score(y_val, y_val_pred), f1_score(y_val, y_val_pred))
print(confusion_matrix(y_val, y_val_pred))
model = SVC(C=2, gamma=0.4, kernel='rbf' ).fit(X, df.target)
y_test_pred = model.predict(X_test)
sub_df ... | Natural Language Processing with Disaster Tweets |
7,928,811 | class CrossEntropyLabelSmooth(nn.Module):
def __init__(self, num_classes, epsilon=0.1, use_gpu=True):
super(CrossEntropyLabelSmooth, self ).__init__()
self.num_classes = num_classes
self.epsilon = epsilon
self.use_gpu = use_gpu
self.logsoftmax = nn.LogSoftmax(dim=1)
def forward(self, inputs, targets):
log_probs = ... | X_all = pd.concat([df["text_preprocessed"], df_test["text_preprocessed"]])
tfidf = TfidfVectorizer(stop_words = 'english')
tfidf.fit(X_all)
del X_all
X = tfidf.transform(df["text_preprocessed"])
X_test = tfidf.transform(df_test["text_preprocessed"])
X_train, X_val, y_train, y_val = train_test_split(X, df["target"]... | Natural Language Processing with Disaster Tweets |
7,928,811 | def fastrcnn_loss(class_logits, box_regression, labels, regression_targets):
labels = torch.cat(labels, dim=0)
regression_targets = torch.cat(regression_targets, dim=0)
labal_smooth_loss = CrossEntropyLabelSmooth(2)
classification_loss = labal_smooth_loss(class_logits, labels)
sampled_pos_inds_subset = torch.nonz... | model = SVC(C=2, gamma=0.4, kernel='rbf' ).fit(X_train, y_train)
y_val_pred = model.predict(X_val)
print(accuracy_score(y_val, y_val_pred), f1_score(y_val, y_val_pred))
print(confusion_matrix(y_val, y_val_pred))
model = SVC(C=2, gamma=0.4, kernel='rbf' ).fit(X, df.target)
y_test_pred = model.predict(X_test)
sub_df ... | Natural Language Processing with Disaster Tweets |
7,928,811 | def fpn_backbone_269(pretrained, norm_layer=misc_nn_ops.FrozenBatchNorm2d, trainable_layers=3):
print(f'
Pretarined is {pretrained}')
backbone = resnest269e(pretrained=pretrained)
assert trainable_layers <= 5 and trainable_layers >= 0
layers_to_train = ['layer4', 'layer3', 'layer2', 'layer1', 'conv1'][:trainable_laye... | X_all = pd.concat([df["text_preprocessed"], df_test["text_preprocessed"]])
tfidf = TfidfVectorizer(stop_words = 'english', min_df=10)
tfidf.fit(X_all)
del X_all
X = tfidf.transform(df["text_preprocessed"])
X_test = tfidf.transform(df_test["text_preprocessed"])
X_train, X_val, y_train, y_val = train_test_split(X, d... | Natural Language Processing with Disaster Tweets |
7,928,811 | def fpn_backbone_101(pretrained, norm_layer=misc_nn_ops.FrozenBatchNorm2d, trainable_layers=3):
print(f'
Pretarined is {pretrained}')
backbone = resnest101e(pretrained=pretrained)
assert trainable_layers <= 5 and trainable_layers >= 0
layers_to_train = ['layer4', 'layer3', 'layer2', 'layer1', 'conv1'][:trainable_laye... | model = SVC(C=2, gamma=0.4, kernel='rbf' ).fit(X_train, y_train)
y_val_pred = model.predict(X_val)
print(accuracy_score(y_val, y_val_pred), f1_score(y_val, y_val_pred))
print(confusion_matrix(y_val, y_val_pred))
model = SVC(C=2, gamma=0.4, kernel='rbf' ).fit(X, df.target)
y_test_pred = model.predict(X_test)
sub_df ... | Natural Language Processing with Disaster Tweets |
7,928,811 | def load_net_269(checkpoint_path):
model = WheatDetector_269()
checkpoint=torch.load(checkpoint_path)
model.load_state_dict(checkpoint['model_state_dict'])
del checkpoint
gc.collect()
model.eval() ;
return model.cuda()
def load_net_101(checkpoint_path):
model = WheatDetector_101()
checkpoint=torch.load(checkpoint_pat... | X_all = pd.concat([df["text_preprocessed"], df_test["text_preprocessed"]])
tfidf = TfidfVectorizer(stop_words = 'english', min_df=5, ngram_range=(1, 3))
tfidf.fit(X_all)
del X_all
X = tfidf.transform(df["text_preprocessed"])
X_test = tfidf.transform(df_test["text_preprocessed"])
X_train, X_val, y_train, y_val = tra... | Natural Language Processing with Disaster Tweets |
7,928,811 | class BaseWheatTTA:
image_size = 1024
def augment(self, image):
raise NotImplementedError
def batch_augment(self, images):
raise NotImplementedError
def deaugment_boxes(self, boxes):
raise NotImplementedError
class TTAHorizontalFlip(BaseWheatTTA):
def augment(self, image):
return image.flip(1)
def batch_augment(se... | model = SVC(C=2, gamma=0.4, kernel='rbf' ).fit(X_train, y_train)
y_val_pred = model.predict(X_val)
print(accuracy_score(y_val, y_val_pred), f1_score(y_val, y_val_pred))
print(confusion_matrix(y_val, y_val_pred))
model = SVC(C=2, gamma=0.4, kernel='rbf' ).fit(X, df.target)
y_test_pred = model.predict(X_test)
sub_df ... | Natural Language Processing with Disaster Tweets |
7,928,811 | def process_det(index, outputs, score_threshold=0.5):
boxes = outputs[index]['boxes'].data.cpu().numpy()
scores = outputs[index]['scores'].data.cpu().numpy()
boxes =(boxes ).clip(min=0, max=1023 ).astype(int)
indexes = np.where(scores>score_threshold)
boxes = boxes[indexes]
scores = scores[indexes]
return boxes, scor... | X_all = pd.concat([df["tokens"], df_test["tokens"]] ).reset_index(drop=True)
print(len(X_all))
mydict = Dictionary(X_all)
corpus = [mydict.doc2bow(text)for text in X_all]
tf_model = TfidfModel(corpus)
corpus_tf = tf_model[corpus]
print(len(corpus_tf))
lsi_model = LsiModel(corpus_tf, id2word=mydict, num_topics=200 ) | Natural Language Processing with Disaster Tweets |
7,928,811 | tta_transforms = []
for tta_combination in product([TTAHorizontalFlip() , None],
[TTAVerticalFlip() , None],
[TTARotate90() , None]):
tta_transforms.append(TTACompose([tta_transform for tta_transform in tta_combination if tta_transform]))<categorify> | mydict.num_docs | Natural Language Processing with Disaster Tweets |
7,928,811 | def make_tta_predictions(images,net, score_threshold=0.1):
with torch.no_grad() :
images = torch.stack(images ).float().cuda()
predictions = []
for tta_transform in tta_transforms:
result = []
outputs = net(tta_transform.batch_augment(images.clone()))
for i, image in enumerate(images):
boxes = outputs[i]['boxes'].data.... | for i, word in enumerate(mydict.items()):
print(word)
if i > 9:
break | Natural Language Processing with Disaster Tweets |
7,928,811 | fold1 = {}
for images, image_ids in data_loader:
predictions = make_tta_predictions(images,models[0])
for i, image in enumerate(images):
boxes, scores, labels = run_wbf(predictions, image_index=i)
image_id = image_ids[i]
fold1[image_id] = [boxes, scores, labels]
print('
Completed')
fold2 = {}
for images, image_ids i... | lsi_model.num_topics | Natural Language Processing with Disaster Tweets |
7,928,811 | fold5 = {}
for images, image_ids in data_loader:
predictions = make_tta_predictions(images,models[4])
for i, image in enumerate(images):
boxes, scores, labels = run_wbf(predictions, image_index=i)
image_id = image_ids[i]
fold5[image_id] = [boxes, scores, labels]
print('
Completed')
fold6 = {}
for images, image_ids i... | def transform(df):
corpus = [mydict.doc2bow(text)for text in df]
corpus = tf_model[corpus]
corpus = lsi_model[corpus]
return corpus
print(len(df["tokens"]))
X = transform(df["tokens"])
print(len(X))
X_test = transform(df_test["tokens"])
X_train, X_val, y_train, y_val = train_test_split(X, df["target"], test_size=0.1,... | Natural Language Processing with Disaster Tweets |
7,928,811 | def run_last_wbf(model1,model2,model3,model4,
model5,model6,model7,model8,model9,model10,
iou_thr=0.5,skip_box_thr=0.43):
box1,scores1,labels1 = model1
box2,scores2,labels2 = model2
box3,scores3,labels3 = model3
box4,scores4,labels4 = model4
box1 = box1/1023
box2 = box2/1023
box3 = box3/1023
box4 = box4/1023
box5,score... | def make_vec(X, num_top):
matrix = np.zeros(( len(X), num_top))
for i, row in enumerate(X):
matrix[i, list(map(lambda tup: tup[0], row)) ] = list(map(lambda tup: tup[1], row))
return matrix
make_vec(X_train, lsi_model.num_topics ).shape | Natural Language Processing with Disaster Tweets |
7,928,811 | w = {}
for row in range(len(all_path)) :
image_id = all_path[row].split("/")[-1].split(".")[0]
boxes,scores,labels = run_last_wbf(fold1[image_id],fold2[image_id],fold3[image_id],fold4[image_id],
fold5[image_id],fold6[image_id],fold7[image_id],fold8[image_id],fold9[image_id],fold10[image_id])
boxes =(boxes*1023)
index... | def transform(df, tf_model, model):
corpus = [mydict.doc2bow(text)for text in df]
corpus = tf_model[corpus]
corpus = model[corpus]
corpus = make_vec(corpus, model.num_topics)
return corpus
print(len(df["tokens"]))
X = transform(df["tokens"], tf_model, lsi_model)
print(len(X))
X_test = transform(df_test["tokens"], tf_... | Natural Language Processing with Disaster Tweets |
7,928,811 | def get_valid_transforms() :
return A.Compose([
A.Resize(height=512, width=512, p=1.0),
ToTensorV2(p=1.0),
], p=1.0 )<data_type_conversions> | model = SVC(C=2, gamma=0.4, kernel='rbf' ).fit(X_train, y_train)
y_val_pred = model.predict(X_val)
print(accuracy_score(y_val, y_val_pred), f1_score(y_val, y_val_pred))
print(confusion_matrix(y_val, y_val_pred))
model = SVC(C=2, gamma=0.4, kernel='rbf' ).fit(X, df.target)
y_test_pred = model.predict(X_test)
sub_df ... | Natural Language Processing with Disaster Tweets |
7,928,811 | class DatasetRetriever(Dataset):
def __init__(self, image_ids, transforms=None):
super().__init__()
self.image_ids = image_ids
self.transforms = transforms
def __getitem__(self, index: int):
image_id = self.image_ids[index]
image = cv2.imread(f'{DATA_ROOT_PATH}/{image_id}.jpg', cv2.IMREAD_COLOR)
image = cv2.cvtColor(i... | model = RandomForestClassifier(n_estimators=200, max_depth=None, random_state=42, n_jobs=-1 ).fit(X_train, y_train)
y_val_pred = model.predict(X_val)
print(accuracy_score(y_val, y_val_pred), f1_score(y_val, y_val_pred))
print(confusion_matrix(y_val, y_val_pred))
model = RandomForestClassifier(n_estimators=200, max_de... | Natural Language Processing with Disaster Tweets |
7,928,811 | dataset = DatasetRetriever(
image_ids=np.array([path.split('/')[-1][:-4] for path in glob(f'{DATA_ROOT_PATH}/*.jpg')]),
transforms=get_valid_transforms()
)
def collate_fn(batch):
return tuple(zip(*batch))
data_loader = DataLoader(
dataset,
batch_size=4,
shuffle=False,
num_workers=2,
drop_last=False,
collate_fn=coll... | lda_model = LdaModel(corpus_tf, id2word=mydict, num_topics=100, dtype=np.float64 ) | Natural Language Processing with Disaster Tweets |
7,928,811 | class CrossEntropyLabelSmooth(nn.Module):
def __init__(self, num_classes, epsilon=0.1, use_gpu=True):
super(CrossEntropyLabelSmooth, self ).__init__()
self.num_classes = num_classes
self.epsilon = epsilon
self.use_gpu = use_gpu
self.logsoftmax = nn.LogSoftmax(dim=1)
def forward(self, inputs, targets):
log_probs = ... | lda_model.num_topics | Natural Language Processing with Disaster Tweets |
7,928,811 | class BaseWheatTTA:
image_size = 512
def augment(self, image):
raise NotImplementedError
def batch_augment(self, images):
raise NotImplementedError
def deaugment_boxes(self, boxes):
raise NotImplementedError
class TTAHorizontalFlip(BaseWheatTTA):
def augment(self, image):
return image.flip(1)
def batch_augment(sel... | print(len(df["tokens"]))
X = transform(df["tokens"], tf_model, lda_model)
print(len(X))
X_test = transform(df_test["tokens"], tf_model, lda_model)
X_train, X_val, y_train, y_val = train_test_split(X, df["target"], test_size=0.1, random_state=42 ) | Natural Language Processing with Disaster Tweets |
7,928,811 | def make_tta_predictions(images,net, score_threshold=0.1):
with torch.no_grad() :
images = torch.stack(images ).float().cuda()
predictions = []
for tta_transform in tta_transforms:
result = []
outputs = net(tta_transform.batch_augment(images.clone()))
for i, image in enumerate(images):
boxes = outputs[i]['boxes'].data.... | model = SVC(C=2, gamma=0.4, kernel='rbf' ).fit(X_train, y_train)
y_val_pred = model.predict(X_val)
print(accuracy_score(y_val, y_val_pred), f1_score(y_val, y_val_pred))
print(confusion_matrix(y_val, y_val_pred))
model = SVC(C=2, gamma=0.4, kernel='rbf' ).fit(X, df.target)
y_test_pred = model.predict(X_test)
sub_df ... | Natural Language Processing with Disaster Tweets |
7,928,811 | fold0 = {}
for images, image_ids in data_loader:
predictions = make_tta_predictions(images,models[0])
for i, image in enumerate(images):
boxes, scores, labels = run_wbf(predictions, image_index=i)
image_id = image_ids[i]
fold0[image_id] = [boxes, scores, labels]
print('
Completed')
fold1 = {}
for images, image_ids i... | model = RandomForestClassifier(n_estimators=200, max_depth=None, random_state=42, n_jobs=-1 ).fit(X_train, y_train)
y_val_pred = model.predict(X_val)
print(accuracy_score(y_val, y_val_pred), f1_score(y_val, y_val_pred))
print(confusion_matrix(y_val, y_val_pred))
model = RandomForestClassifier(n_estimators=200, max_de... | Natural Language Processing with Disaster Tweets |
7,928,811 | def run_last_wbf(model1,model2,model3,model4,model5,
iou_thr=0.5,skip_box_thr=0.43):
box1,scores1,labels1 = model1
box2,scores2,labels2 = model2
box3,scores3,labels3 = model3
box4,scores4,labels4 = model4
box5,scores5,labels5 = model5
box1 = box1/1023
box2 = box2/1023
box3 = box3/1023
box4 = box4/1023
box5 = box5/1023
... | print(len(df["tokens"]))
X_1 = transform(df["tokens"], tf_model, lda_model)
X_2 = transform(df["tokens"], tf_model, lsi_model)
X = np.hstack(( X_1, X_2))
print(len(X))
X_test = transform(df_test["tokens"], tf_model, lda_model)
X_1 = transform(df_test["tokens"], tf_model, lda_model)
X_2 = transform(df_test["tokens"]... | Natural Language Processing with Disaster Tweets |
7,928,811 | def load_net(checkpoint_path):
config = get_efficientdet_config('tf_efficientdet_d5')
net = EfficientDet(config, pretrained_backbone=False)
config.num_classes = 1
config.image_size=512
net.class_net = HeadNet(config, num_outputs=config.num_classes, norm_kwargs=dict(eps=.001, momentum=.01))
checkpoint = torch.load(che... | model = SVC(C=2, gamma=0.4, kernel='rbf' ).fit(X_train, y_train)
y_val_pred = model.predict(X_val)
print(accuracy_score(y_val, y_val_pred), f1_score(y_val, y_val_pred))
print(confusion_matrix(y_val, y_val_pred))
model = SVC(C=2, gamma=0.4, kernel='rbf' ).fit(X, df.target)
y_test_pred = model.predict(X_test)
sub_df ... | Natural Language Processing with Disaster Tweets |
7,928,811 | DATA_ROOT_PATH = '.. /input/global-wheat-detection/test/'
class TestDatasetRetriever(Dataset):
def __init__(self, image_ids, transforms=None):
super().__init__()
self.image_ids = image_ids
self.transforms = transforms
def __getitem__(self, index: int):
image_id = self.image_ids[index]
image = cv2.imread(f'{DATA_ROOT_PA... | model = RandomForestClassifier(n_estimators=200, max_depth=None, random_state=42, n_jobs=-1 ).fit(X_train, y_train)
y_val_pred = model.predict(X_val)
print(accuracy_score(y_val, y_val_pred), f1_score(y_val, y_val_pred))
print(confusion_matrix(y_val, y_val_pred))
model = RandomForestClassifier(n_estimators=200, max_de... | Natural Language Processing with Disaster Tweets |
7,928,811 | def run_wbf(predictions, image_index, image_size=512, iou_thr=0.432, skip_box_thr=0.397, weights=None):
boxes = [(prediction[image_index]['boxes']/(image_size-1)).tolist() for prediction in predictions]
scores = [prediction[image_index]['scores'].tolist() for prediction in predictions]
labels = [np.ones(prediction[imag... | X_all = pd.concat([df["tokens"], df_test["tokens"]] ).reset_index(drop=True)
model_w2v = Word2Vec(sentences=X_all, size=50, window=3, min_count=1, workers=-1)
del X_all | Natural Language Processing with Disaster Tweets |
7,928,811 | y={}
for images, image_ids in data_loader:
predictions = make_predictions(images)
for i, image in enumerate(images):
boxes, scores, labels = run_wbf(predictions, image_index=i,)
boxes =(boxes*2)
y[image_ids[i]] = [boxes,scores,labels]<load_pretrained> | vector = model_w2v.wv['armageddon']
print(vector ) | Natural Language Processing with Disaster Tweets |
7,928,811 | models = [
load_net('.. /input/efficientdetd52/0.bin'),
load_net('.. /input/efficientdetd52/1.bin'),
load_net('.. /input/efficientdetd52/2.bin'),
load_net('.. /input/efficientdetd52/3.bin'),
load_net('.. /input/efficientdetd52/4.bin'),
load_net('.. /input/kaggleeffnet/plabel_model/last-checkpoint1.bin'),
load_net('.. /... | print(len(df["tokens"]))
X = [np.mean([model_w2v.wv[text] for text in texts], axis=0)for texts in df["tokens"]]
X = np.array(X)
print(len(X))
X_test = [np.mean([model_w2v.wv[text] for text in texts], axis=0)if len(texts)!= 0 else np.zeros(50)for texts in df_test["tokens"]]
X_test = np.array(X_test)
X_test.shape | Natural Language Processing with Disaster Tweets |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.