Max Reimann
commited on
Commit
β’
11a70dd
1
Parent(s):
dc6a058
add page for xdog prediction
Browse files- images/apdrawing/img_1585.png +3 -0
- images/apdrawing/img_1592.png +3 -0
- images/apdrawing/img_1594.png +3 -0
- images/apdrawing/img_1600.png +3 -0
- images/apdrawing/img_1607.png +3 -0
- images/apdrawing/img_1616.png +3 -0
- pages/3_π§_Predict_Portrait_xDoG.py +194 -0
- pages/{3_π_Readme.py β 4_π_Readme.py} +0 -0
- requirements.txt +2 -1
images/apdrawing/img_1585.png
ADDED
Git LFS Details
|
images/apdrawing/img_1592.png
ADDED
Git LFS Details
|
images/apdrawing/img_1594.png
ADDED
Git LFS Details
|
images/apdrawing/img_1600.png
ADDED
Git LFS Details
|
images/apdrawing/img_1607.png
ADDED
Git LFS Details
|
images/apdrawing/img_1616.png
ADDED
Git LFS Details
|
pages/3_π§_Predict_Portrait_xDoG.py
ADDED
@@ -0,0 +1,194 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import argparse
|
2 |
+
import base64
|
3 |
+
from io import BytesIO
|
4 |
+
from pathlib import Path
|
5 |
+
import os
|
6 |
+
import shutil
|
7 |
+
import sys
|
8 |
+
import time
|
9 |
+
|
10 |
+
import numpy as np
|
11 |
+
import torch.nn.functional as F
|
12 |
+
import torch
|
13 |
+
import streamlit as st
|
14 |
+
from st_click_detector import click_detector
|
15 |
+
|
16 |
+
from matplotlib import pyplot as plt
|
17 |
+
from mpl_toolkits.axes_grid1 import make_axes_locatable
|
18 |
+
from torchvision.transforms import ToPILImage, Compose, ToTensor, Normalize
|
19 |
+
from PIL import Image
|
20 |
+
|
21 |
+
from huggingface_hub import hf_hub_download
|
22 |
+
|
23 |
+
|
24 |
+
PACKAGE_PARENT = '..'
|
25 |
+
WISE_DIR = '../wise/'
|
26 |
+
SCRIPT_DIR = os.path.dirname(os.path.realpath(os.path.join(os.getcwd(), os.path.expanduser(__file__))))
|
27 |
+
sys.path.append(os.path.normpath(os.path.join(SCRIPT_DIR, PACKAGE_PARENT)))
|
28 |
+
sys.path.append(os.path.normpath(os.path.join(SCRIPT_DIR, WISE_DIR)))
|
29 |
+
|
30 |
+
|
31 |
+
from local_ppn.options.test_options import TestOptions
|
32 |
+
from local_ppn.models import create_model
|
33 |
+
|
34 |
+
|
35 |
+
|
36 |
+
class CustomOpts(TestOptions):
|
37 |
+
|
38 |
+
def remove_options(self, parser, options):
|
39 |
+
for option in options:
|
40 |
+
for action in parser._actions:
|
41 |
+
print(action)
|
42 |
+
if vars(action)['option_strings'][0] == option:
|
43 |
+
parser._handle_conflict_resolve(None,[(option,action)])
|
44 |
+
break
|
45 |
+
|
46 |
+
def initialize(self, parser):
|
47 |
+
parser = super(CustomOpts, self).initialize(parser)
|
48 |
+
self.remove_options(parser, ["--dataroot"])
|
49 |
+
return parser
|
50 |
+
|
51 |
+
def print_options(self, opt):
|
52 |
+
pass
|
53 |
+
|
54 |
+
def add_predefined_images():
|
55 |
+
images = []
|
56 |
+
for f in os.listdir(os.path.join(SCRIPT_DIR, PACKAGE_PARENT, 'images','apdrawing')):
|
57 |
+
if not f.endswith('.png'):
|
58 |
+
continue
|
59 |
+
AB = Image.open(os.path.join(SCRIPT_DIR, PACKAGE_PARENT, 'images','apdrawing', f)).convert('RGB')
|
60 |
+
# split AB image into A and B
|
61 |
+
w, h = AB.size
|
62 |
+
w2 = int(w / 2)
|
63 |
+
A = AB.crop((0, 0, w2, h))
|
64 |
+
B = AB.crop((w2, 0, w, h))
|
65 |
+
images.append(A)
|
66 |
+
return images
|
67 |
+
|
68 |
+
@st.experimental_singleton
|
69 |
+
def make_model(_unused=None):
|
70 |
+
model_path = hf_hub_download(repo_id="MaxReimann/WISE-APDrawing-XDoG", filename="apdrawing_xdog_ppn_conv.pth")
|
71 |
+
os.makedirs(os.path.join(SCRIPT_DIR, PACKAGE_PARENT, "trained_models", "ours_apdrawing"), exist_ok=True)
|
72 |
+
shutil.copy2(model_path, os.path.join(SCRIPT_DIR, PACKAGE_PARENT, "trained_models", "ours_apdrawing", "latest_net_G.pth"))
|
73 |
+
|
74 |
+
opt = CustomOpts().parse() # get test options
|
75 |
+
# hard-code some parameters for test
|
76 |
+
opt.num_threads = 0 # test code only supports num_threads = 0
|
77 |
+
opt.batch_size = 1 # test code only supports batch_size = 1
|
78 |
+
# opt.serial_batches = True # disable data shuffling; comment this line if results on randomly chosen images are needed.
|
79 |
+
opt.no_flip = True # no flip; comment this line if results on flipped images are needed.
|
80 |
+
opt.display_id = -1 # no visdom display; the test code saves the results to a HTML file.
|
81 |
+
opt.dataroot ="null"
|
82 |
+
opt.direction = "BtoA"
|
83 |
+
opt.model = "pix2pix"
|
84 |
+
opt.ppnG = "our_xdog"
|
85 |
+
opt.name = "ours_apdrawing"
|
86 |
+
opt.netG = "resnet_9blocks"
|
87 |
+
opt.no_dropout = True
|
88 |
+
opt.norm = "batch"
|
89 |
+
opt.load_size = 576
|
90 |
+
opt.crop_size = 512
|
91 |
+
opt.eval = False
|
92 |
+
model = create_model(opt) # create a model given opt.model and other options
|
93 |
+
model.setup(opt) # regular setup: load and print networks; create schedulers
|
94 |
+
if opt.eval:
|
95 |
+
model.eval()
|
96 |
+
|
97 |
+
|
98 |
+
return model, opt
|
99 |
+
|
100 |
+
def predict(image):
|
101 |
+
model, opt = make_model()
|
102 |
+
t = Compose([
|
103 |
+
ToTensor(),
|
104 |
+
Normalize(mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5))
|
105 |
+
])
|
106 |
+
inp = image.resize((opt.crop_size, opt.crop_size), resample=Image.BICUBIC)
|
107 |
+
inp = t(inp).unsqueeze(0).cuda()
|
108 |
+
x = model.netG.module.ppn_part_forward(inp)
|
109 |
+
|
110 |
+
output = model.netG.module.conv_part_forward(x)
|
111 |
+
out_img = ToPILImage()(output.squeeze(0))
|
112 |
+
return out_img
|
113 |
+
|
114 |
+
|
115 |
+
|
116 |
+
st.title("xDoG+CNN Portrait Drawing ")
|
117 |
+
|
118 |
+
images = add_predefined_images()
|
119 |
+
|
120 |
+
html_code = '<div class="column" style="display: flex; flex-wrap: wrap; padding: 0 4px;">'
|
121 |
+
for i, image in enumerate(images):
|
122 |
+
buffered = BytesIO()
|
123 |
+
image.save(buffered, format="JPEG")
|
124 |
+
encoded = base64.b64encode(buffered.getvalue()).decode()
|
125 |
+
html_code += f"<a href='#' id='{i}' style='padding: 0px 5px'><img height='120px' style='margin-top: 8px;' src='data:image/jpeg;base64,{encoded}'></a>"
|
126 |
+
html_code += "</div>"
|
127 |
+
clicked = click_detector(html_code)
|
128 |
+
|
129 |
+
uploaded_im = st.file_uploader(f"OR: Load portrait:", type=["png", "jpg"], )
|
130 |
+
if uploaded_im is not None:
|
131 |
+
img = Image.open(uploaded_im)
|
132 |
+
img = img.convert('RGB')
|
133 |
+
buffered = BytesIO()
|
134 |
+
img.save(buffered, format="JPEG")
|
135 |
+
|
136 |
+
|
137 |
+
clicked_img = None
|
138 |
+
if clicked:
|
139 |
+
clicked_img = images[int(clicked)]
|
140 |
+
|
141 |
+
sel_img = img if uploaded_im is not None else clicked_img
|
142 |
+
if sel_img:
|
143 |
+
result_container = st.container()
|
144 |
+
coll1, coll2 = result_container.columns([3,2])
|
145 |
+
coll1.header("Result")
|
146 |
+
coll2.header("Global Edits")
|
147 |
+
|
148 |
+
model, opt = make_model()
|
149 |
+
t = Compose([
|
150 |
+
ToTensor(),
|
151 |
+
Normalize(mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5))
|
152 |
+
])
|
153 |
+
inp = sel_img.resize((opt.crop_size, opt.crop_size), resample=Image.BICUBIC)
|
154 |
+
inp = t(inp).unsqueeze(0).cuda()
|
155 |
+
# vp = model.netG.module.ppn_part_forward(inp)
|
156 |
+
|
157 |
+
vp = model.netG.module.predict_parameters(inp)
|
158 |
+
inp = (inp * 0.5) + 0.5
|
159 |
+
|
160 |
+
effect = model.netG.module.apply_visual_effect.effect
|
161 |
+
|
162 |
+
with coll2:
|
163 |
+
# ("blackness", "contour", "strokeWidth", "details", "saturation", "contrast", "brightness")
|
164 |
+
show_params_names = ["strokeWidth", "blackness", "contours"]
|
165 |
+
display_means = []
|
166 |
+
params_mapping = {"strokeWidth": ['strokeWidth'], 'blackness': ["blackness"], "contours": [ "details", "contour"]}
|
167 |
+
def create_slider(name):
|
168 |
+
params = params_mapping[name] if name in params_mapping else [name]
|
169 |
+
means = [torch.mean(vp[:, effect.vpd.name2idx[n]]).item() for n in params]
|
170 |
+
display_mean = float(np.average(means) + 0.5)
|
171 |
+
display_means.append(display_mean)
|
172 |
+
slider = st.slider(f"Mean {name}: ", 0.0, 1.0, value=display_mean, step=0.05)
|
173 |
+
for i, param_name in enumerate(params):
|
174 |
+
vp[:, effect.vpd.name2idx[param_name]] += slider - (means[i]+ 0.5)
|
175 |
+
# vp.clamp_(-0.5, 0.5)
|
176 |
+
# pass
|
177 |
+
|
178 |
+
for name in show_params_names:
|
179 |
+
create_slider(name)
|
180 |
+
|
181 |
+
x = model.netG.module.apply_visual_effect(inp, vp)
|
182 |
+
x = (x - 0.5) / 0.5
|
183 |
+
|
184 |
+
only_x_dog = st.checkbox('only xdog', value=False, help='if checked, use only ppn+xdog, else use ppn+xdog+post-processing cnn')
|
185 |
+
if only_x_dog:
|
186 |
+
output = x[:,0].repeat(1,3,1,1)
|
187 |
+
print('shape output', output.shape)
|
188 |
+
else:
|
189 |
+
output = model.netG.module.conv_part_forward(x)
|
190 |
+
|
191 |
+
out_img = ToPILImage()(output.squeeze(0))
|
192 |
+
output = out_img.resize((320,320), resample=Image.BICUBIC)
|
193 |
+
with coll1:
|
194 |
+
st.image(output)
|
pages/{3_π_Readme.py β 4_π_Readme.py}
RENAMED
File without changes
|
requirements.txt
CHANGED
@@ -10,4 +10,5 @@ streamlit==1.10.0
|
|
10 |
streamlit_drawable_canvas==0.8.0
|
11 |
streamlit_extras==0.1.5
|
12 |
st_click_detector
|
13 |
-
scipy
|
|
|
|
10 |
streamlit_drawable_canvas==0.8.0
|
11 |
streamlit_extras==0.1.5
|
12 |
st_click_detector
|
13 |
+
scipy
|
14 |
+
huggingface_hub
|