Yuktha-Majella

## Load_Resnet50
from fastai.vision import *
from fastai.vision.models import resnet50

## Data_augmentation
tfms = get_transforms(do_flip=True, flip_vert=True, max_rotate=50, max_lighting=0.1, max_warp=0 )
data = ImageDataBunch.from_df('/content/drive/MyDrive/CV_Vehicle_classification/train_data/images', train, ds_tfms=tfms, label_delim= None, valid_pct=0.2, fn_col=0, label_col=1 , size=299,bs=64).normalize(imagenet_stats)

## Model_train
t_cnn1 = cnn_learner(data, resnet50, pretrained=True, metrics=[accuracy])
t_cnn1.fit_one_cycle(5)

## model_train_unfreeze
t_cnn1.unfreeze()
t_cnn1.fit_one_cycle(8)

## Finding_LR
t_cnn1.lr_find()
t_cnn1.recorder.plot()


## freeze_model
t_cnn1.fit_one_cycle(10,max_lr=slice(1e-5, 1e-4))
t_cnn1.freeze()
t_cnn1.export('/content/drive/MyDrive/CV_Vehicle_classification/model/Bmodel_fastai_resnet50.h5')

## Predication
test_data = ImageList.from_df(test, cols=['image_names'], path='/content/drive/MyDrive/CV_Vehicle_classification/train_data/images')
t_rn50 = load_learner('/content/drive/MyDrive/CV_Vehicle_classification/model/', 'Bmodel_fastai_resnet50.h5', test = test_data)
y_trn50 = t_rn50.TTA(ds_type = DatasetType.Test)
preds = y_trn50[0].argmax(-1)

## create_dict
import gensim
from gensim import corpora

text1 = ["""Gensim is a free open-source Python library for representing documents as semantic vectors,
           as efficiently and painlessly as possible. Gensim is designed
           to process raw, unstructured digital texts using unsupervised machine learning algorithms."""]

tokens1 = [[item for item in line.split()] for line in text1]
g_dict1 = corpora.Dictionary(tokens1)

## create_dict_textfile
from gensim.utils import simple_preprocess
from gensim import corpora

text2 = open('sample_text.txt', encoding ='utf-8')

tokens2 =[]
for line in text2.read().split('.'):
  tokens2.append(simple_preprocess(line, deacc = True))

g_dict2 = corpora.Dictionary(tokens2)

## dict_update
g_dict1.add_documents(tokens2)

print("The dictionary has: " +str(len(g_dict1)) + " tokens\n")
print(g_dict1.token2id)
	from fastai.vision import *
	from fastai.vision.models import resnet50
	tfms = get_transforms(do_flip=True, flip_vert=True, max_rotate=50, max_lighting=0.1, max_warp=0 )
	data = ImageDataBunch.from_df('/content/drive/MyDrive/CV_Vehicle_classification/train_data/images', train, ds_tfms=tfms, label_delim= None, valid_pct=0.2, fn_col=0, label_col=1 , size=299,bs=64).normalize(imagenet_stats)
	t_cnn1 = cnn_learner(data, resnet50, pretrained=True, metrics=[accuracy])
	t_cnn1.fit_one_cycle(5)
	t_cnn1.fit_one_cycle(10,max_lr=slice(1e-5, 1e-4))
	t_cnn1.freeze()
	t_cnn1.export('/content/drive/MyDrive/CV_Vehicle_classification/model/Bmodel_fastai_resnet50.h5')
	test_data = ImageList.from_df(test, cols=['image_names'], path='/content/drive/MyDrive/CV_Vehicle_classification/train_data/images')
	t_rn50 = load_learner('/content/drive/MyDrive/CV_Vehicle_classification/model/', 'Bmodel_fastai_resnet50.h5', test = test_data)
	y_trn50 = t_rn50.TTA(ds_type = DatasetType.Test)
	preds = y_trn50[0].argmax(-1)
	import gensim
	from gensim import corpora

	text1 = ["""Gensim is a free open-source Python library for representing documents as semantic vectors,
	as efficiently and painlessly as possible. Gensim is designed
	to process raw, unstructured digital texts using unsupervised machine learning algorithms."""]

	tokens1 = [[item for item in line.split()] for line in text1]
	g_dict1 = corpora.Dictionary(tokens1)
	from gensim.utils import simple_preprocess
	from gensim import corpora

	text2 = open('sample_text.txt', encoding ='utf-8')

	tokens2 =[]
	for line in text2.read().split('.'):
	tokens2.append(simple_preprocess(line, deacc = True))

	g_dict2 = corpora.Dictionary(tokens2)
	g_dict1.add_documents(tokens2)

	print("The dictionary has: " +str(len(g_dict1)) + " tokens\n")
	print(g_dict1.token2id)