from io import BytesIO
from datetime import datetime
from flask import Flask, request, render_template
from random import random, choice
from lib import data_util
from lib.config import params_setup
from lib.googlenet import GoogLeNet
app = Flask(__name__)
# model
# scope_name, label_size = '17flowers', 17
# scope_name, label_size = '17portraits', 9
args = params_setup()
gnet = GoogLeNet(args=args)
#---------------------------
# Server
#---------------------------
@app.route('/', methods=['GET'])
def guess():
url = request.args.get('url', '')
if url:
X = url2sample(url)
probs = gnet.predict([X])[0]
cnt = int(sum([math.exp(i + 4) * probs[i] for i in range(len(probs))]))
probs = [(i, round(100 * p, 1)) for i, p in enumerate(probs)]
else:
cnt, probs = None, None
from __future__ import division, print_function, absolute_import
from lib import data_util
from lib.config import params_setup
from lib.googlenet import GoogLeNet
from datetime import datetime
import pickle, gzip
import numpy as np
import tflearn.datasets.oxflower17 as oxflower17
#-------------------------------
# Training
#-------------------------------
# scope_name, label_size = '17flowers', 17
# scope_name, label_size = '17portraits', 9
args = params_setup()
gnet = GoogLeNet(args=args) #img_size=227, label_size=label_size, gpu_memory_fraction=0.4, scope_name=scope_name)
pkl_files = gnet.get_data(dirname=args.model_name, down_sampling=args.down_sampling)
epoch = 0
while True:
for f in pkl_files:
X, Y = pickle.load(gzip.open(f, 'rb'))
gnet.fit(X, Y, n_epoch=10)
print('[pkl_files] done with %s @ %s' % (f, datetime.now()))
epoch += 1
# print("[Finish] all pkl_files been trained %i times." % epoch)
from lib.config import params_setup
from lib.googlenet import GoogLeNet
from datetime import datetime
import tensorflow as tf
import tflearn
import pickle, gzip
import numpy as np
import tflearn.datasets.oxflower17 as oxflower17
# -------------------------------
# Training
# -------------------------------
args = params_setup()
gnet = GoogLeNet(args=args)
# img_size=227, label_size=label_size, gpu_memory_fraction=0.4, scope_name=scope_name)
pkl_files = gnet.get_data(dirname=args.model_name,
down_sampling=args.down_sampling)
epoch = 0
while True:
for f in pkl_files:
X, Y = pickle.load(gzip.open(f, 'rb'))
with tf.device('/device:GPU:0'):
gnet.fit(X, Y, n_epoch=10)
print('[pkl_files] done with %s @ %s' % (f, datetime.now()))
epoch += 1
print("[Finish] all pkl_files been trained %i times." % epoch)
import pickle, gzip
import numpy as np
import tflearn.datasets.oxflower17 as oxflower17
# import AI Vision train service module
from dnn_train import train_service
#-------------------------------
# Training
#-------------------------------
# init AI Vision train service
train_service = train_service.TrainService()
args = params_setup()
gnet = GoogLeNet(args, train_service)
# go to pre-processing stage
train_service.sendStatusMessagePreproccess()
# go to training stage
train_service.sendStatusMessageTrain()
print(pkl_files)
for f in pkl_files:
X, Y = pickle.load(gzip.open(f, 'rb'))
gnet.fit(X, Y, n_epoch=100)
gnet.save()
print('[pkl_files] done with %s @ %s' % (f, datetime.now()))
# go to complete stage
train_service.sendStatusMessageComplete()
import numpy as np
from PIL import Image
from scipy import misc
from io import BytesIO
from datetime import datetime
from flask import Flask, request, render_template
from random import random, choice
from lib import data_util
from lib.config import params_setup
from lib.googlenet import GoogLeNet
# model
args = params_setup()
gnet = GoogLeNet(args=args)
directory_names = list(set(glob.glob(os.path.join("images","tiny-imagenet-200","jpg", "*"))\
).difference(set(glob.glob(os.path.join("images","tiny-imagenet-200","jgp","*.*",)))))
total_count = 0.
correct_count = 0.
# len(directory_names)
for i in range(len(directory_names)):
imgs_in_folder = glob.glob(
os.path.join(directory_names[i], "images", "*.JPEG"))
for j in range(400, len(imgs_in_folder)):
cur_img = imgs_in_folder[j]
# img = imread(cur_img)
# img = load_image(s)
# img = Image.open(cur_img)
# # img = resize_image(img, 227, 227)