def tab(filename="info"):
'''
分页组件
'''
tab = Tab()
tab.add(liquid_data_precision("111"), "前日留存")
tab.add(bar_datazoom_slider("222"), "分端留存")
tab.add(table_base("222"), "明细")
tab.add(line_markpoint("222"), "人群留存")
tab.add(pie_rosetype("222"), "城市")
tab.add(grid_mutil_yaxis("222"), "新增留存")
tab.render_notebook()
tab.render('{}.html'.format(filename))
def test_tab_render_notebook():
tab = Tab()
tab.add(_create_line(), "line-example")
tab.add(_create_bar(), "bar-example")
tab.add(_create_table(), "table-example")
html = tab.render_notebook().__html__()
assert_in("City name", html)
def visualize_layer(model, image, layer_name, layer_max_image=32, jupyter=True, path='visualize_layer.html'):
"""network layer visualize.
Args:
model: a tf.keras model or keras model.
image: a image array with shape (1, height, width, channel), height and width are necessary.
layer_name: a list of model layers name.
layer_max_image: every layer max plot images.
jupyter: if plot in jupyter, default True.
path: if jupyter is False, result save a html file.
Returns:
A pyecharts polt object.
"""
if tf.io.gfile.exists('feature_map'):
tf.io.gfile.rmtree('feature_map')
tf.io.gfile.makedirs('feature_map')
if np.ndim(image)==2:
image = tf.expand_dims(tf.expand_dims(image, 0), -1)
elif np.ndim(image)==3:
image = tf.expand_dims(image, 0)
elif np.ndim(image)==4:
pass
else:
raise ValueError('image shape is error.')
temp_model = tf.keras.backend.function(model.inputs, [i.output for i in model.layers if i.name in layer_name])
temp_name = [i.name for i in model.layers if i.name in layer_name]
result = temp_model(image)
images_per_row = 16
count = 0
name_dict = {}
tab = Tab()
for feature, name in zip(result, temp_name):
if feature.ndim==4:
if feature.shape[-1]==3:
display_grid = feature[0,:,:,:].astype('uint8')
else:
n_features = feature.shape[-1] if feature.shape[-1]<layer_max_image else layer_max_image
size = feature.shape[1]
n_cols = int(np.ceil(n_features/images_per_row))
display_grid = np.ones((size * n_cols, images_per_row * size),dtype=np.uint8)*255
for col in range(n_cols):
for row in range(images_per_row):
if (col+1)*(row+1)>n_features:
break
channel_image = feature[0, :, :, col * images_per_row + row]
channel_image -= channel_image.mean()
channel_image /= channel_image.std()
channel_image *= 64
channel_image += 128
channel_image = np.clip(channel_image, 0, 255).astype('uint8')
display_grid[col * size : (col + 1) * size, row * size : (row + 1) * size] = channel_image
display_grid = np.expand_dims(display_grid, axis=-1)
name_dict[name] = f'./feature_map/{count}.png'
ImageAug(display_grid).save_image(name_dict[name])
tab.add(image_base(name_dict[name], name, 'shape='+str(feature.shape)), name)
count += 1
return tab.render_notebook() if jupyter else tab.render(path)
def visualize_heatmaps(model, image, layer_name, jupyter=True, path='visualize_heatmaps.html'):
"""network layer visualize.
Args:
model: a tf.keras model or keras model.
image: a image array with shape (1, height, width, channel), height and width are necessary.
layer_name: a list of model layers name.
jupyter: if plot in jupyter, default True.
path: if jupyter is False, result save a html file.
Returns:
A pyecharts polt object.
"""
if tf.io.gfile.exists('feature_map'):
tf.io.gfile.rmtree('feature_map')
tf.io.gfile.makedirs('feature_map')
if np.ndim(image)==2:
image = tf.expand_dims(tf.expand_dims(image, 0), -1)
elif np.ndim(image)==3:
image = tf.expand_dims(image, 0)
elif np.ndim(image)==4:
pass
else:
raise ValueError('image shape is error.')
temp_model = tf.keras.backend.function(model.inputs, [i.output for i in model.layers if i.name in layer_name])
temp_name = [i.name for i in model.layers if i.name in layer_name]
result = temp_model(image)
images_per_row = 16
count = 0
name_dict = {}
tab = Tab()
for feature, name in zip(result, temp_name):
if feature.ndim==4:
if model.get_layer(name).__class__.__name__=='InputLayer':
out = np.squeeze(image, 0).astype('uint8')
else:
out = tf.image.resize(tf.expand_dims(tf.squeeze(tf.reduce_sum(tf.abs(feature), axis=-1)), axis=-1), (image.shape[1], image.shape[2]))
out = 255-tf.cast(out/tf.reduce_max(out)*255., tf.uint8)
# print(np.squeeze(image, 0).shape)
out = addWeighted(applyColorMap(out.numpy(), COLORMAP_JET), 0.7, np.squeeze(image, 0).astype('uint8'), 0.3, 0)
# out = np.expand_dims(out, axis=-1)
name_dict[name] = f'./feature_map/{count}.png'
ImageAug(out).save_image(name_dict[name])
tab.add(image_base(name_dict[name], name, 'shape='+str(feature.shape)), name)
count += 1
return tab.render_notebook() if jupyter else tab.render(path)
is_show=True)),
))
return c #返回数据
effectscatter_symbol().render_notebook()
tab = Tab() #添加tab
tab.add(teachers_map(), "特殊教育专任教师数")
tab.add(schools_map(), "各省特殊教育学校数")
tab.add(recruit_students_map(), "各省特殊教育招生数")
tab.add(all(), "教育支出与特殊教育在校学生详情")
tab.add(finacial_education_map(), "地方财政教育支出")
tab.add(four_compare(), "四个不同省份对比")
tab.add(effectscatter_symbol(), "四川特殊教育基本情况")
tab.render_notebook()
@app.route('/', endpoint='r1', redirect_to='/index')
def root():
return render_template('index.html')
@app.route('/index')
def index():
return render_template('index.html')
@app.route('/local_expense')
def local_expense():
c = tab = Tab()
title_opts=opts.TitleOpts(title="Grid-多 Y 轴示例"),
tooltip_opts=opts.TooltipOpts(trigger="axis", axis_pointer_type="cross"),
)
)
line = (
Line()
.add_xaxis(x_data)
.add_yaxis(
"平均温度",
[2.0, 2.2, 3.3, 4.5, 6.3, 10.2, 20.3, 23.4, 23.0, 16.5, 12.0, 6.2],
yaxis_index=2,
color="#675bba",
label_opts=opts.LabelOpts(is_show=False),
)
)
bar.overlap(line)
return Grid().add(
bar, opts.GridOpts(pos_left="5%", pos_right="20%"), is_control_axis_index=True
)
tab = Tab()
tab.add(bar_datazoom_slider(), "bar-example")
tab.add(line_markpoint(), "line-example")
tab.add(pie_rosetype(), "pie-example")
tab.add(grid_mutil_yaxis(), "grid-example")
put_html(tab.render_notebook())