def single_get_fc8(arr):
with tf.Session() as sess:
images = tf.placeholder(tf.float32, [1,224,224,3])
# 这里images的占位形状为[一次喂入的图片张数，图片长，图片宽，图片通道信息红绿蓝]
vgg = vgg16.Vgg16()# 实例化vgg，运行了Vgg16类的初始化函数__ini__
vgg.forward(images)
probability = sess.run(vgg.fc8, feed_dict={images: arr})
return probability
def get_fc8(img_path):
img_ready = utils.load_image(img_path)# 读入待判图，并将其处理成[None,224,224,3]的形状
with tf.Session() as sess:
images = tf.placeholder(tf.float32, [1,224,224,3])
# 这里images的占位形状为[一次喂入的图片张数，图片长，图片宽，图片通道信息红绿蓝]
vgg = vgg16.Vgg16()# 实例化vgg，运行了Vgg16类的初始化函数__ini__
vgg.forward(images)
probability = sess.run(vgg.fc8, feed_dict={images: img_ready})
return probability
def batch_get_fc8(img_ready_array):
# img_ready_list = []
# for img_path in img_path_list:
# img_ready_list.append(utils.load_image(img_path))# 读入待判图，并将其处理成[1,224,224,3]的形状
# img_ready_array = np.array(img_ready_list)
print(img_ready_array.shape)
with tf.Session() as sess:
images = tf.placeholder(tf.float32, [None,224,224,3])
# 这里images的占位形状为[一次喂入的图片张数，图片长，图片宽，图片通道信息红绿蓝]
vgg = vgg16.Vgg16()# 实例化vgg，运行了Vgg16类的初始化函数__ini__
vgg.forward(images)
fc8_feas = sess.run(vgg.fc8, feed_dict={images: img_ready_array})
return fc8_feas
def pred_class(img_path):
img_ready = utils.load_image(img_path)# 读入待判图，并将其处理成[1,224,224,3]的形状
# fig = plt.figure(u"Top-5 prediction")# 准备一张图，把运行的结果可视化。
with tf.Session() as sess:
images = tf.placeholder(tf.float32, [1,224,224,3])
# 这里images的占位形状为[一次喂入的图片张数，图片长，图片宽，图片通道信息红绿蓝]
vgg = vgg16.Vgg16()# 实例化vgg，运行了Vgg16类的初始化函数__ini__
vgg.forward(images)
probability = sess.run(vgg.prob, feed_dict={images: img_ready})
# probability存着1000个概率，对应1000中每个类别的概率，索引key是labels中的健
# 用probability[0][i]遍历每个概率值
top5 = np.argsort(probability[0])[-1:-6:-1]
# np.argsort表示对列表从小到大排序，返回索引值。这里的top5存着probability列表中5个最高概率的索引
print("top:", top5)
values = []# 用来存probability中元素的值，存概率
bar_label = []# 用来存标签字典中对应的值，存名称
for n, i in enumerate(top5):
# print("n:", n)
# print("i:", i)
values.append(probability[0][i])
bar_label.append(labels[i])
print(i, ":", labels[i], "----", utils.percent(probability[0][i]))
'''
ax = fig.add_subplot(111)
# fig.add_subplot(数 数 数)，111表示：包含1行，包含1列，当前是第1个
ax.bar(range(len(values)), values, tick_label=bar_label, width=0.5, fc='g')
# ax.bar(bar的个数，bar的值，每个bar的名字，bar宽，bar色)
ax.set_ylabel(u'probabilityit')
# ax.set_ylabel(" ")设置第一列的y轴名字
ax.set_title(u'Top-5')
# ax.set_title(" ")设置子图名字
for a, b in zip(range(len(values)), values):
ax.text(a, b+0.0005, utils.percent(b), ha='center', va='bottom', fontsize=7)
# ax.text(文字x坐标，文字y坐标，文字内容，ha='center'水平方向位置，va='bottom'垂直方向位置，字号)
plt.show()
'''
return values[0], bar_label[0]
def fig_vectorize():
# train = pd.read_pickle(os.path.join(os.path.dirname(__file__), '..', 'output', 'train.pd'))# 读入train
# train['figs_reshaped_arr'] = train['fig_path_list'].apply(lambda list: figs_reshaped_arr(list))
# pd.to_pickle(train, os.path.join(os.path.dirname(__file__), '..', 'output', 'train_reshapedfig.pd'))
# test = pd.read_pickle(os.path.join(os.path.dirname(__file__), '..', 'output', 'test.pd'))
# test['figs_reshaped_arr'] = test['fig_path_list'].apply(lambda list: figs_reshaped_arr(list))
# pd.to_pickle(test, os.path.join(os.path.dirname(__file__), '..', 'output', 'test_reshapedfig.pd'))
with tf.Session() as sess:
images = tf.placeholder(tf.float32, [None,224,224,3])
# 这里images的占位形状为[一次喂入的图片张数，图片长，图片宽，图片通道信息红绿蓝]
vgg = vgg16.Vgg16()# 实例化vgg，运行了Vgg16类的初始化函数__ini__
vgg.forward(images)
## train========
## 因为文件太大，所以分开存储
# chunk_num = 1500
# for i in range(chunk_num):
# train_temp = pd.read_pickle(os.path.join(os.path.dirname(__file__), '..', 'output', 'train_reshapedfig.pd'))
# chunk_size_train = math.ceil(len(train_temp) / chunk_num)
# train = train_temp.iloc[i*chunk_size_train: (i+1)*chunk_size_train]
# del train_temp
# gc.collect()
#
# arr_list = []
# for arr in train['figs_reshaped_arr']:
# if type(arr)!=float:
# arr_list.append(arr)
# del train
# gc.collect()
# reshaped_arr_ensemble = np.concatenate(arr_list, axis=0)
# print(reshaped_arr_ensemble.shape)
# print('train reshaped fig arr ensemble finished')
# del arr_list
# gc.collect()
#
# fig_feas = sess.run(vgg.fc8, feed_dict={images: reshaped_arr_ensemble})
# np.save(os.path.join(os.path.dirname(__file__), '..', 'output', 'midFile', 'train_figFC8_%d.npy'%(i)), fig_feas)
# print('train feature finished %d'%(i))
# del reshaped_arr_ensemble, fig_feas
# gc.collect()
## test=========
chunk_num = 950
for i in range(chunk_num):
test_temp = pd.read_pickle(os.path.join(os.path.dirname(__file__), '..', 'output', 'test_reshapedfig.pd'))
chunk_size_test = math.ceil(len(test_temp) / chunk_num)
test = test_temp.iloc[i*chunk_size_test: (i+1)*chunk_size_test]
del test_temp
gc.collect()
arr_list = []
for arr in test['figs_reshaped_arr']:
if type(arr)!=float:
arr_list.append(arr)
del test
gc.collect()
reshaped_arr_ensemble = np.concatenate(arr_list, axis=0)
print(reshaped_arr_ensemble.shape)
print('test reshaped fig arr ensemble finished')
del arr_list
gc.collect()
fig_feas = sess.run(vgg.fc8, feed_dict={images: reshaped_arr_ensemble})
np.save(os.path.join(os.path.dirname(__file__), '..', 'output', 'midFile', 'test_figFC8_%d.npy'%(i)), fig_feas)
print('test feature finished %d'%(i))
del reshaped_arr_ensemble, fig_feas
gc.collect()