def start_predict(args):
# ~ print(args)
options = {
'img_path': args.image_path,
'weights_path': args.weights_path,
'result_dir': args.result_image_dir
}
predict(options)
def main():
# Measures total program runtime by collecting start time
start_time = time()
# Collect training parameter from the console interface
in_arg = get_input_predict()
# Check consistency of the input parameters
control_input_args_predict(in_arg.checkpoint, in_arg.mapping_file, in_arg.arch, in_arg.gpu, int(in_arg.topk))
# Choose betwen GPU and CPU
if in_arg.gpu[:1].lower()== 'y':
device = 'cuda:0'
else:
device = 'cpu'
print(f"\n Running on '{device}' device...\n")
# Predict the class of the entered flower image with the help of the chosen trained model
probs, classes = predict(in_arg.flower_image, in_arg.checkpoint, in_arg.arch, device, int(in_arg.topk))
# Determine class with higest probability
i_max = 0
prob_max =0
for i in range(len(probs)):
if probs[i] > prob_max:
i_max = i
prob_max = probs[i]
# Mapping of Class value to category name
with open(in_arg.mapping_file, 'r') as f:
cat_to_name = json.load(f)
classes_name = []
for classe in classes:
for key, value in cat_to_name.items():
if str(key) == str(classe) :
classes_name.append(str(value))
# Printing results of the prediction
print(f"\nThe network used is for this classification is : {in_arg.checkpoint} \n")
print(f"Top 5 classes are: {classes} \n ")
print(f"Top 5 classes names are: {classes_name} \n ")
print(f"Top 5 classes probabilities are: {probs} \n ")
print(f"The classe with highest probability is {classes_name[i_max]} with a probability of {probs[i_max]} \n ")
# Measure total program runtime by collecting end time
end_time =time()
# Computes overall runtime in seconds & prints it in hh:mm:ss format
tot_time = end_time-start_time
print("\n** Total Elapsed Runtime:",
str(int((tot_time/3600)))+":"+str(int((tot_time%3600)/60))+":"
+str(int((tot_time%3600)%60)) )
def main():
# Parse the arguments passed by the user
parsed_arguments = arg_parser()
# Check if both checkpoint and image file exist
if os.path.isfile(parsed_arguments.path_to_image) and os.path.isfile(
parsed_arguments.checkpoint):
# Make sure that the model is available on the used device
device = torch.device("cuda:0" if (
torch.cuda.is_available() and parsed_arguments.gpu) else "cpu")
print(device)
# Load a model from a checkpoint
loaded_model = model_functions.load_checkpoint(
parsed_arguments.checkpoint)
loaded_model.to(device)
#print(loaded_model)
processed_image = data_functions.process_image(
parsed_arguments.path_to_image)
# Don't allow the top_K to be lower than 1
topk = parsed_arguments.top_K
if topk < 1:
topk = 1
probs, classes = model_functions.predict(processed_image, loaded_model,
parsed_arguments.gpu, topk)
if os.path.isfile(parsed_arguments.category_names):
# Create a mapping from category label to category name
cat_to_name = data_functions.label_mapping(
parsed_arguments.category_names)
for item, prob in zip(classes, probs):
print("The probability that image {} is a '{}' is {}%".format(
parsed_arguments.path_to_image, cat_to_name[str(item)],
prob))
else:
print(
'Failed to get category to name mapping! File does not exist: %s',
parsed_argument.category_names)
print('Probabilities {}'.format(probs))
print('Classes {}'.format(classes))
else:
print(
"Either image or the checkpoint file doesn't exist. Please check the parameters."
)
return
'Type your tweet (up to 280 characters) to check positivity or \'end\' if you\'re done: '
)
while userText != 'end':
assert len(userText) < 280
if input('Select the model to use: Tf-Idf (1) or Count Vectorizer (2): '
) == '1':
model = TfIdf_Model
vectorizer = TfIdf_Vectorizer
modelLabel = 'Tf-Idf'
else:
model = CountVect_Model
vectorizer = CountVect_Vect
modelLabel = 'Count Vectorizer'
tweetObj = MF.predict([userText], vectorizer, model, silence=True)
tweetColor = np.ones((3, 3)) * tweetObj[0][2]
tweetColor = tweetColor if tweetObj[0][1] == 'Positive' else -1 * tweetColor
tweetColor = (tweetColor + 1) / 2
fig, ax = plt.subplots()
im = ax.imshow(tweetColor, cmap=cm.get_cmap('RdYlGn'), vmax=1.0, vmin=0.0)
ax.axis('off')
if len(userText) > 20:
fontSize = 15
else:
fontSize = -3 / (2 * len(userText)) + 30
plt.text(1,
1,
userText,
help='class_to_name json file')
parser.add_argument('--device', type=str, default='cuda', help='GPU or CPU')
arguments = parser.parse_args()
# Load in a mapping from category label to category name
class_to_name_map = utility_functions.load_json(arguments.json)
# Load pretrained network
model = model_functions.get_checkpoint(arguments.checkpoint)
print(model)
checkpoint = torch.load(arguments.checkpoint)
# Scales, crops, and normalizes a PIL image for the PyTorch model; returns a Numpy array
image = utility_functions.process_image(arguments.image_path)
# Display image
processing_functions.imshow(image)
# Highest k probabilities and the indices of those probabilities corresponding to the classes (converted to the actual class labels)
probabilities, classes = model_functions.predict(arguments.image_path, model,
arguments.topk,
arguments.device)
print(probabilities)
print(classes)
# Display the image along with the top 5 classes
processing_functions.display_image(arguments.image_path, class_to_name_map,
classes)
return parser.parse_args()
arguments = get_input_args()
#getting the arguments ready to use
if arguments.gpu:
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
else:
device = "cpu"
#load checkpoint
loaded_checkpoint = load_checkpoint(arguments.checkpoint, device)
criterion = loaded_checkpoint['criterion']
optimizer = loaded_checkpoint['optimizer']
epochs = loaded_checkpoint['epochs']
model = loaded_checkpoint['model']
#load class names mapping
mapping_of_categories = load_classes_json(arguments.category_names)
# #predict probs
top_outputs = predict(arguments.path, model, mapping_of_categories,arguments.top_k, device)
# print("Command Line Arguments:\n path =", arguments.path,
# "\n checkpoint =", arguments.checkpoint,
# "\n category_names =", arguments.category_names,
# "\n gpu =", arguments.gpu,
# "\n top_k =", arguments.top_k)
print(top_outputs)
def classify():
image = None
#----------------- Uploading User Tree image from URL or File -------------------
st.title(
'Please provide an image - will be identified as pepper tree, willow tree, or neither.'
)
#----------------- Side Bar Options for User Image ------------------------------
st.title("Upload Options")
input_method = st.radio("Options", ('File Upload', 'URL'))
flag = 0
if input_method == 'File Upload':
user_upload_image = st.file_uploader("Upload a picture of Tree",
type=['png', 'jpeg', 'jpg'])
if user_upload_image is not None:
file_details = {
"FileName": user_upload_image.name,
"FileType": user_upload_image.type,
"FileSize": user_upload_image.size
}
# st.write(file_details)
flag = 1
if flag == 1:
image_source = user_upload_image.name
image = Image.open(user_upload_image)
st.image(
image,
caption=user_upload_image.name + ' ' +
user_upload_image.type + ' ' + str(user_upload_image.size) +
' bytes',
width=300,
use_column_width=True,
)
# Image from URL
if input_method == 'URL':
image_url = st.text_area("Enter the complete Url",
key="user_url_choice")
image_url_status = st.button('Upload')
if image_url_status:
image_source = image_url
image = Image.open(urllib.request.urlopen(image_url))
st.image(
image,
caption=str(image),
width=300,
use_column_width=True,
)
else:
st.warning('click on upload')
#---------------------- Choosing Classification Method ---------------------------
st.title('Choose the model for Analysis')
model_selected = st.radio("Options", ['Pre Trained Model'])
# model_selected = st.sidebar.radio(
# "Options",
# ('Pre Trained Model', 'CNN Model', 'FFN Model', 'Random Forest', 'Logistic Regression', 'Ensemble'),0)
if model_selected == 'Pre Trained Model':
model_selected2 = st.selectbox("Choose the Pretrained Model",
['VGG16'])
# model_selected2 = st.sidebar.selectbox("Choose the Pretrained Model",
# ['VGG16','PTM2','PTM2','PTM2','PTM2','PTM2','PTM2','PTM2'])
if model_selected2 == 'VGG16' and image != None:
# note that the predict function returns top_probabilities, top_classes
model = load_checkpoint(
'/home/kate/data_and_models/ai_models/vgg16-tree-3class-model.pth')
probs, classes = predict(image, model)
st.title('Tree Classification Results')
with open('tree_to_name.json', 'r') as f:
tree_to_name = json.load(f)
tree_names = [tree_to_name[i] for i in classes]
chart_data = pd.DataFrame(data=[probs], columns=tree_names)
# st.write("chart_data type ", type(chart_data))
if (chart_data["willow tree"][0]) > 0.5:
tree_detected = "Willow Tree"
elif (chart_data["pepper tree"][0]) > 0.5:
tree_detected = "Pepper Tree"
else:
tree_detected = "Not a Pepper Tree or a Willow Tree"
st.write('The image is: ', tree_detected)
st.write('Percentage confidence in the image identification ',
chart_data)
st.bar_chart(chart_data)
style_image = cv2.resize(style_image,
(style_scaled_width, content_scaled_height))
# Draw content and style images
st.image([content_image, style_image],
caption=['Content Image', 'Style Image'],
width=None)
# Create options dictionary for predict function
options = dict()
options['img_path'] = content_image
options['weights_path'] = weights_path
options['result_dir'] = None
# Predict and display result
predicted_img = predict(options, write_result=False)
st.write("Result image")
st.image(predicted_img[:, :, ::-1], width=500)
# Download results
result = Image.fromarray(predicted_img)
def get_image_download_link(img):
"""Generates a link allowing the PIL image to be downloaded
in: PIL image
out: href string
"""
buffered = BytesIO()
img.save(buffered, format="JPEG")
img_str = base64.b64encode(buffered.getvalue()).decode()
def check_mentions(api, since_id, time_now):
new_since_id = since_id
print('new check')
for tweet in tweepy.Cursor(api.mentions_timeline,
since_id=since_id).items():
user_handle = tweet.user.screen_name
new_since_id = max(tweet.id, new_since_id)
if time_now > tweet.created_at:
continue
print('new tweet since id', new_since_id)
if tweet.in_reply_to_status_id is not None:
continue
if not tweet.user.following:
tweet.user.follow()
# Parse user input
tweet_input = tweet.text.replace('@musicpsychic2', '')
separator = tweet_input.find(',')
song_name = tweet_input[:separator].strip()
artist_name = tweet_input[separator + 1:].strip()
print(artist_name)
print(song_name)
# Retrieve song dictionary
song_dict = read_song_dict(artist_name, song_name)
# print(song_dict['valence'])
if song_dict == {}:
api.update_status(
status="@" + user_handle + " I can't find the song " +
song_name + " or the artist " + artist_name +
" please check your spelling and the list of artists I have data on in my pinned tweet.",
in_reply_to_status_id=tweet.id)
continue
# retrieve top 5 words for song
rank_df = get_song_tf_idf(song_dict, 5)
# retrieve positive vs. negative prediction from predict function
song_pred = predict(song_dict)
# create response string
# Change response to say: artist talks about blank, blank, ... and blank to convey generally ___ emotions
response = ' ' + artist_name + ' uses words like'
word_list = list(rank_df.index)
for i in range(len(word_list)):
if i == len(word_list) - 1:
response += ' and ' + word_list[i]
else:
response += ' ' + word_list[i] + ','
response += ' to convey '
if song_pred == 1:
response += "generally positive emotions in " + song_name + '.'
else:
response += "generally negative emotions in " + song_name + '.'
api.update_status(status="@" + user_handle + response,
in_reply_to_status_id=tweet.id)
return new_since_id
help='class_to_name json file')
parser.add_argument('--gpu', type=str, default='cuda', help='GPU or CPU')
arguments = parser.parse_args()
# Load in a mapping from category label to category name
class_to_name_dict = processing_functions.load_json(arguments.json)
# Load pretrained network
model = model_functions.load_checkpoint(arguments.checkpoint)
print(model)
checkpoint = torch.load(arguments.checkpoint)
# Scales, crops, and normalizes a PIL image for the PyTorch model; returns a Numpy array
image = processing_functions.process_image(arguments.image_dir,
checkpoint['hidden_layer_units'])
# Display image
processing_functions.imshow(image)
# Highest k probabilities and the indices of those probabilities corresponding to the classes (converted to the actual class labels)
probabilities, classes = model_functions.predict(arguments.image_dir, model,
arguments.topk, arguments.gpu)
print(probabilities)
print(classes)
# Display the image along with the top 5 classes
processing_functions.display_image(arguments.image_dir, class_to_name_dict,
classes)
type=int)
parser.add_argument('--category_names',
dest="category_names",
action="store",
default='cat_to_name.json')
parse = parser.parse_args()
outputs = parse.top_k
processor = parse.gpu
input_img = parse.img
chkt_path = parse.checkpoint
model_load = model_functions.load_checkpoint(chkt_path)
model_functions.load_checkpoint(filepath)
# cpu
device = torch.device("cpu")
# gpu
if parse.gpu:
device = torch.device("cuda:0")
with open(category_names) as json_file:
cat_to_name = json.load(json_file)
top_probabilities, top_classes = model_functions.predict(
img_path, model, outputs)
print(top_probabilities)
print(top_classes)
parser.add_argument('image', action='store')
parser.add_argument('checkpoint', action='store')
parser.add_argument('--cat_names', action='store', default='cat_to_name.json')
parser.add_argument('--gpu', action='store', type=bool, default=True)
parser.add_argument('--top_k', action='store', type=int, default=5)
args = parser.parse_args()
##### PREDICTION
### Load model checkpoint and dictionary
#train_dataset, trainloader = load_train('flowers/train')
model, epochs, learn, optim_state, optimizer, criterion, classes_d = load_checkpoint(
args.checkpoint)
label_dict = get_labels(args.cat_names)
#model.class_to_idx = train_dataset.class_to_idx
### Configure device to use - gpu or cpu
device = choose_device(args.gpu)
### Load and process image, run forward propagation to obtain probabilities and classes for top_k predictions
image = process_image(args.image)
probabilities, classes = predict(image, model, device, classes_d, k=args.top_k)
### Print result
print('The top {} predictions of the model are the following:'.format(
args.top_k))
for p, c, r in zip(probabilities, classes, range(1, args.top_k + 1)):
print('{}. {}, p = {}'.format(r, label_dict[str(c)], round(p, 3)))
# Load dictionary mapping labels to category names
cat_to_name = 'cat_to_name.json'
label_dict = label_mapping(cat_to_name)
# Run the image through the model and obtain the topk predictions
topk = 5
# Identify the device on which the prediction will be performed
#GPU = arg_dict['gpu']
GPU = False
if GPU:
device = torch.device('cuda')
else:
device = torch.device('cpu')
probs, classes = predict(image, model, topk, device)
# Create list of labels for the topk classes
labels = []
for cls in classes:
labels.append(label_dict[cls])
# Plot the cropped input image and the topk classes with probabilities
fig, (ax1,ax2) = plt.subplots(2,1,figsize=(4,8))
# Undo transposition of dimensions required by PyTorch
image = image.transpose((1, 2, 0))
# Undo preprocessing
mean = np.array([0.485, 0.456, 0.406])
std = np.array([0.229, 0.224, 0.225])
image = std * image + mean
vectorizer = CountVect_Vect
modelLabel = 'Count Vectorizer'
print('Retrieving 100 of the most recent tweets of user: '******'data'][0]['name'], '...')
userTweets, tweetDates = user_tweets.main(userDat['data'][0]['id'],
tweetCount)
# strip punctuation
userTweets = [
''.join(c for c in s if c not in string.punctuation)
for s in userTweets
]
tweetWords = [word for line in userTweets for word in line.split()]
wordCounter = Counter({})
wordCounter = wordCounter + Counter(tweetWords)
tweetObj = MF.predict(userTweets, vectorizer, model, silence=True)
stopwords = set(STOPWORDS)
redlist = []
greenlist = []
# count word frequency
wordCounterDict1 = dict(wordCounter)
wordCounterDict1 = dict(
sorted(wordCounterDict1.items(),
key=lambda item: item[1],
reverse=True))
listDict1 = list(wordCounterDict1.keys())
wordCounterDict = wordCounterDict1
listDict = list(wordCounterDict.keys())
from model_functions import build_model, load_checkpoint, predict
import argparse
parser = argparse.ArgumentParser(
description='use trained model to predict class of a image ')
parser.add_argument('path_to_image', help='the path of image')
parser.add_argument('model_directory',
help='the directory of model that has been trained')
parser.add_argument('--top_K',
type=int,
default=3,
help='return the top K most possible classes')
parser.add_argument('--category_names',
default='',
help='output the actual class name')
parser.add_argument('--gpu',
action='store_true',
default=False,
help='the training mode')
args = parser.parse_args()
model, other_info = load_checkpoint(args.model_directory)
image = process_image(args.path_to_image)
classes, probs = predict(image,
model,
topk=args.top_K,
category_name=args.category_names,
gpu_truth=args.gpu,
other_info=other_info)
print(classes)
print(probs)
args = vars(parser.parse_args())
args_dict = dict(
filter(lambda elem: (elem[1] != None) and (elem[1] != False),
args.items()))
kwargs_mapping = {
k: v
for (k, v) in args_dict.items() if k in ['category_names']
}
cat_to_name = utility.load_class_mapping(**kwargs_mapping)
model, optimizer = utility.load_checkpoint(args_dict['checkpoint'])
kwargs_predict = {
k: v
for (k, v) in args_dict.items() if k in ['top_k', 'gpu']
}
probs, classes = model_functions.predict(args_dict['image_path'], model,
**kwargs_predict)
classes_names = [cat_to_name[cl] for cl in classes]
predictions = zip(classes_names, probs)
print("Prediction:")
[
print("{} with a probability of {:.3}".format(cl, p))
for cl, p in predictions
]
## Test: python predict.py "flowers/test/1/image_06743.jpg" checkpoint_1.pth --top_k 3
