-
Notifications
You must be signed in to change notification settings - Fork 0
/
ui.py
200 lines (140 loc) · 6.25 KB
/
ui.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
import dash
from dash import dcc, html, Input, Output
import plotly.express as px
from vae_api import VAE_API
import torch
import torchvision
from torchvision import transforms
from vae_api import VAE_API
from model_config import num_epochs, z_dim
import os
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
dataset = torchvision.datasets.MNIST(root='../../data',
train=True,
transform=transforms.ToTensor(),
download=True)
vae_api = VAE_API(os.path.join(os.getcwd(), "checkpoints/"), dataset, batch_size=512)
import dash_bootstrap_components as dbc
IMAGE_1 = None
IMAGE_2 = None
df = vae_api.generate_scatterplot_dataframe(num_epochs)
GRAPH = px.scatter(df, x="x", y="y", color="labels", title="tSNE embeddings of latent space vectors", template="plotly_dark")
RADIO_BUTTONS = ["Image 1", "Image 2"]
INITIAL_RADIO_SELECTION = RADIO_BUTTONS[0]
RADIO_SELECTION = RADIO_BUTTONS[0]
####
## Variables for latent space visualization
LATENT_SPACE_DIM = z_dim
LAST_SELECTED_VECTOR = [0,0,0,0,0,0,0,0,0,0]
LAST_RECONSTRUCTEd_VECTOR = None
latent_vector_1 = None
latent_vector_2 = None
# app = Dash(__name__)
app = dash.Dash(external_stylesheets=[dbc.themes.CYBORG])
# create assets dir if it doesn't already exist
assets_directory = os.path.join(os.getcwd(), "assets/")
if not os.path.exists(assets_directory):
os.makedirs(assets_directory)
latent_space_inputs = [html.H2("Sampled Latent Vector")]
for i in range(LATENT_SPACE_DIM):
latent_space_inputs.append(
dcc.Input(id=f"latent_input_{i}", type = "number", placeholder = f"x{i}", value="0", readOnly=True, style={"color": "#ffffff", "background-color": "#080000"}))#}))
latent_space_inputs.append(html.Br())
latent_space_inputs.append(
html.Button("Reconstruct last selected latent vector", id = "copy_latent_vector")
)
latent_space_inputs.append(html.Br())
app.layout = html.Div([
html.H1("VAE Visualizer", style={"text-align":"center"}),
dcc.Graph(id="scatter-plot", figure=GRAPH),
html.Div([
html.Div([
# latent-space interpolation div
html.Div([
dcc.RadioItems(RADIO_BUTTONS, INITIAL_RADIO_SELECTION, id="radio_button", inputStyle ={"margin-left":"50px"}),
html.H4(f"Image Being Edited: {INITIAL_RADIO_SELECTION}", id="radio_info"),
html.H4( f"{RADIO_BUTTONS[0]}", id="point1"),
html.Img(src='', id="image1", height=200, width=200),
html.H4(f"{RADIO_BUTTONS[1]}", id="point2"),
html.Img(src="", id="image2", height=200, width=200), html.Br(),
html.Button("Interpolate", id = "interpolate"),
], style={'padding': 10, 'flex': 1}),
# separate div for the actual gif
html.Div([
html.H2("Interpolation GIF"),
html.Img(src="", id="interpolation-gif", height=300, width=300, style={"vertical-align": "middle"})], style={'padding': 10, 'flex': 1})
], style={'padding': 10, 'flex': 1, 'display': 'flex', 'flex-direction': 'row'}),
# latent-space reconstruction div
html.Div([html.Div(latent_space_inputs, style={'padding': 10, 'flex': 1}), html.Div([html.H2("Reconstructed Image"), html.Img(src='', id="generated_img", height=300, width=300)], style={'padding': 10, 'flex': 1})], style={'padding': 10, 'flex': 1, 'display': 'flex', 'flex-direction': 'row'})
], style={'display': 'flex', 'flex-direction': 'row'})
])
# change point selection radio button (for latent-space interpolation gifs)
@app.callback(
Output("radio_info", "children"),
Input("radio_button", "value"))
def update_radio_selection(option):
print("callback1")
global RADIO_SELECTION
RADIO_SELECTION = option
print(f"RADIO_SELECTION: {RADIO_SELECTION}")
return f"Image Being Edited: {RADIO_SELECTION}"
@app.callback(
Output("interpolation-gif", "src"),
# Input("num-steps", "value"),
Input("interpolate", "n_clicks"))
def get_interpolation_gif(n_clicks):
global latent_vector_1, latent_vector_2
if latent_vector_1 == None or latent_vector_2 == None:
return ""
else:
# print(latent_vector_1)
# print(latent_vector_2)
return vae_api.generate_iterpolation_gif(latent_vector1=latent_vector_1, latent_vector_2=latent_vector_2, num_steps=100)
@app.callback(
Output("latent_input_0", "value"),
Output("latent_input_1", "value"),
Output("latent_input_2", "value"),
Output("latent_input_3", "value"),
Output("latent_input_4", "value"),
Output("latent_input_5", "value"),
Output("latent_input_6", "value"),
Output("latent_input_7", "value"),
Output("latent_input_8", "value"),
Output("latent_input_9", "value"),
Output('generated_img', 'src'),
Input('copy_latent_vector', 'n_clicks'),
)
def update_latent_vector(n_clicks):
global LAST_SELECTED_VECTOR
if LAST_SELECTED_VECTOR == None:
LAST_SELECTED_VECTOR = [0]*LATENT_SPACE_DIM
return LAST_SELECTED_VECTOR + [vae_api.generate_image(LAST_SELECTED_VECTOR)]
# choose point on scatterplot for latent-space interpolation
@app.callback(
Output('image1', 'src'),
Output('image2', 'src'),
Input('scatter-plot', 'clickData'))
def display_click_data(clickData):
print("callback4")
global IMAGE_1, IMAGE_2, RADIO_SELECTION, df, LAST_SELECTED_VECTOR, latent_vector_1, latent_vector_2
# print(clickData)
# print(RADIO_SELECTION)
if clickData == None:
return IMAGE_1, IMAGE_2
# print(clickData["points"])
if RADIO_SELECTION == RADIO_BUTTONS[0]:
# print("editing data point 1")
# IMAGE_1 = os.path.join(os.getcwd(), f'assets/{clickData["points"][0]["pointIndex"]}.png')
index = clickData["points"][0]["pointIndex"]
IMAGE_1 = f'./assets/{clickData["points"][0]["pointIndex"]}.png'
LAST_SELECTED_VECTOR = df.loc[index]["z"]
latent_vector_1 = LAST_SELECTED_VECTOR
else:
IMAGE_2 = os.path.join(os.getcwd(), f'assets/{clickData["points"][0]["pointIndex"]}.png')
index = clickData["points"][0]["pointIndex"]
IMAGE_2 = f'./assets/{clickData["points"][0]["pointIndex"]}.png'
LAST_SELECTED_VECTOR = df.loc[index]["z"]
latent_vector_2 = LAST_SELECTED_VECTOR
return IMAGE_1, IMAGE_2
if __name__ == "__main__":
app.run_server(debug=True)