forked from RogerJzx/MeshMultiViewRender
/
main.py
277 lines (230 loc) · 9.43 KB
/
main.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
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
import numpy as np
from glumpy import gl, app, gloo, glm
from utils import *
import cv2
## glumpy setup
# Color vertex shader
#-------------------------------------------------------------------------------
_color_vertex_code = """
uniform mat4 u_mv;
uniform mat4 u_nm;
uniform mat4 u_mvp;
uniform vec3 u_light_eye_pos;
attribute vec3 a_position;
attribute vec3 a_normal;
attribute vec3 a_color;
attribute vec2 a_texcoord;
varying vec3 v_color;
varying vec2 v_texcoord;
varying vec3 v_eye_pos;
varying vec3 v_L;
varying vec3 v_normal;
void main() {
gl_Position = u_mvp * vec4(a_position, 1.0);
v_color = a_color;
v_texcoord = a_texcoord;
v_eye_pos = (u_mv * vec4(a_position, 1.0)).xyz; // Vertex position in eye coords.
v_L = normalize(u_light_eye_pos - v_eye_pos); // Vector to the light
v_normal = normalize(u_nm * vec4(a_normal, 1.0)).xyz; // Normal in eye coords.
}
"""
# Color fragment shader - flat shading
#-------------------------------------------------------------------------------
_color_fragment_flat_code = """
uniform float u_light_ambient_w;
uniform sampler2D u_texture;
uniform int u_use_texture;
varying vec3 v_color;
varying vec2 v_texcoord;
varying vec3 v_eye_pos;
varying vec3 v_L;
void main() {
// Face normal in eye coords.
vec3 face_normal = normalize(cross(dFdx(v_eye_pos), dFdy(v_eye_pos)));
float light_diffuse_w = max(dot(normalize(v_L), normalize(face_normal)), 0.0);
float light_w = u_light_ambient_w + light_diffuse_w;
if(light_w > 1.0) light_w = 1.0;
if(bool(u_use_texture)) {
gl_FragColor = vec4(light_w * texture2D(u_texture, v_texcoord));
}
else {
gl_FragColor = vec4(light_w * v_color, 1.0);
}
}
"""
def render_rgb(model, cts, view, ssaa=4., shape=(640.480),
K=np.array([[550, 0.0, 316.],
[0.0, 540., 244.],
[0.0, 0.0, 1.0]])):
"""
:param model:load ply model mesh
:param view: multi view points
:param ssaa:defalut = 4.0
# Super-sampling anti-aliasing (SSAA)
# https://github.com/vispy/vispy/wiki/Tech.-Antialiasing
# The RGB image is rendered at ssaa_fact times higher resolution and then
# down-sampled to the required resolution.
:param shape: rgb image shape
:param K: camera intrisic matrix
:return: images
"""
assert ({'pts', 'faces'}.issubset(set(model.keys())))
texture_uv = np.zeros((model['pts'].shape[0], 2), np.float32)
if 'colors' in model.keys():
assert (model['pts'].shape[0] == model['colors'].shape[0])
colors = model['colors']
if colors.max() > 1.0:
colors /= 255.0 # Color values are expected in range [0, 1]
vertices_type = [('a_position', np.float32, 3),
('a_color', np.float32, colors.shape[1]),
('a_texcoord', np.float32, 2)]
vertices = np.array(zip(model['pts'],
colors, texture_uv), vertices_type)
V = vertices.view(gloo.VertexBuffer)
I = model['faces'].flatten().astype(np.uint32).view(gloo.IndexBuffer)
program = gloo.Program(_color_vertex_code, _color_fragment_flat_code)
program.bind(V)
yz_flip = np.eye(4, dtype=np.float32)
yz_flip[1, 1], yz_flip[2, 2] = -1, -1
view = yz_flip.dot(view) # OpenCV to OpenGL camera system
view = view.T
proj = compute_calib_proj(K * ssaa, 0, 0, int(shape[0] * ssaa), int(shape[1] * ssaa), 10, 10000)
program['u_light_eye_pos'] = [0, 0, 0] # Camera origin
program['u_light_ambient_w'] = 0.8
program['u_use_texture'] = int(False)
program['u_texture'] = np.zeros((1, 1, 4), np.float32)
model = (np.eye(4, dtype=np.float32)) ## model matrix
mvp = compute_model_view_proj(model, view, proj)
program['u_mvp'] = mvp
window = app.Window(visible=False)
points = []
viwe_port_matrix = np.array([[shape[0]/2, 0, 0, shape[0]/2],
[0, -shape[1]/2, 0, shape[1]/2],
[0,0,0,0],
[0,0,0,1]])
cts = np.concatenate((cts, np.ones((9,1), dtype=cts.dtype)), axis=1)
for i in range(cts.shape[0]):
p = cts[i]
coors = viwe_port_matrix.dot(mvp.T.dot(p))
coors/=coors[-1]
points.append(coors[0])
points.append(coors[1])
global rgb
rgb = None
@window.event
def on_draw(dt):
# window.clear()
global rgb
extent_shape = (int(shape[0] * ssaa), int(shape[1] * ssaa))
# Frame buffer object
color_buf = np.zeros((extent_shape[0], extent_shape[1], 4), np.float32).view(gloo.TextureFloat2D)
depth_buf = np.zeros((extent_shape[0], extent_shape[1]), np.float32).view(gloo.DepthTexture)
fbo = gloo.FrameBuffer(color=color_buf, depth=depth_buf)
fbo.activate()
# OpenGL setup
gl.glEnable(gl.GL_DEPTH_TEST)
gl.glClear(gl.GL_COLOR_BUFFER_BIT | gl.GL_DEPTH_BUFFER_BIT)
gl.glViewport(0, 0, extent_shape[1], extent_shape[0])
gl.glDisable(gl.GL_CULL_FACE)
program.draw(gl.GL_TRIANGLES, I)
rgb = np.zeros((extent_shape[0], extent_shape[1], 4), dtype=np.float32)
gl.glReadPixels(0, 0, extent_shape[1], extent_shape[0], gl.GL_RGBA, gl.GL_FLOAT, rgb)
rgb.shape = extent_shape[0], extent_shape[1], 4
rgb = rgb[::-1, :]
rgb = np.round(rgb[:, :, :3] * 255).astype(np.uint8) # Convert to [0, 255]
import cv2
rgb = cv2.resize(rgb, shape, interpolation=cv2.INTER_AREA)
fbo.deactivate()
app.run(framecount=0)
window.close()
return rgb, points
def sample_views(min_n_views, radius=1,
azimuth_range=(0, 2 * math.pi),
elev_range=(-0.5 * math.pi, 0.5 * math.pi)):
'''
Viewpoint sampling from a view sphere.
:param min_n_views: Minimum required number of views on the whole view sphere.
:param radius: Radius of the view sphere.
:param azimuth_range: Azimuth range from which the viewpoints are sampled.
:param elev_range: Elevation range from which the viewpoints are sampled.
:return: List of views, each represented by a 3x3 rotation matrix and
a 3x1 translation vector.
'''
# Get points on a sphere
if True:
pts, pts_level = hinter_sampling(min_n_views, radius=radius)
else:
pts = fibonacci_sampling(min_n_views + 1, radius=radius)
pts_level = [0 for _ in range(len(pts))]
views = []
for pt in pts:
# Azimuth from (0, 2 * pi)
azimuth = math.atan2(pt[1], pt[0])
if azimuth < 0:
azimuth += 2.0 * math.pi
# Elevation from (-0.5 * pi, 0.5 * pi)
a = np.linalg.norm(pt)
b = np.linalg.norm([pt[0], pt[1], 0])
elev = math.acos(b / a)
if pt[2] < 0:
elev = -elev
# if hemisphere and (pt[2] < 0 or pt[0] < 0 or pt[1] < 0):
if not (azimuth_range[0] <= azimuth <= azimuth_range[1] and
elev_range[0] <= elev <= elev_range[1]):
continue
# Rotation matrix
# The code was adopted from gluLookAt function (uses OpenGL coordinate system):
# [1] http://stackoverflow.com/questions/5717654/glulookat-explanation
# [2] https://www.opengl.org/wiki/GluLookAt_code
f = -np.array(pt) # Forward direction
f /= np.linalg.norm(f)
u = np.array([0.0, 0.0, 1.0]) # Up direction
s = np.cross(f, u) # Side direction
if np.count_nonzero(s) == 0:
# f and u are parallel, i.e. we are looking along or against Z axis
s = np.array([1.0, 0.0, 0.0])
s /= np.linalg.norm(s)
u = np.cross(s, f) # Recompute up
R = np.array([[s[0], s[1], s[2]],
[u[0], u[1], u[2]],
[-f[0], -f[1], -f[2]]])
# Convert from OpenGL to OpenCV coordinate system
R_yz_flip = rotation_matrix(math.pi, [1, 0, 0])[:3, :3]
R = R_yz_flip.dot(R)
# Translation vector
t = -R.dot(np.array(pt).reshape((3, 1)))
views.append({'R': R, 't': t})
return views, pts_level
rgb = None
def generate_rgbs(rgb_path='data/images',
label_path='data/labels',
vis_path='data/views.ply', vis=False):
shape = (640,480)
model = load_ply('data/object.ply')
model['pts'] = model['pts'] * 100. # Scale points to match view points
## 9 contorls points(1 for centriod point, the rest 8 points are vertexs)
control_points = np.array([[0,0,0],[0.2, 0.4, 0.2], [-0.2, 0.4, 0.2], [-0.2,-0.4, 0.2], [0.2,-0.4, 0.2],
[ 0.2,-0.4,-0.2], [ 0.2, 0.4,-0.2], [-0.2, 0.4,-0.2], [-0.2,-0.4,-0.2]])*100. # Scale points to match view points
azimuth_range = (math.pi, 2 * math.pi)
elev_range = (0, 0.5 * math.pi) # (-59, 90) [deg]
# elev_range = (0, math.pi*45/180.)
min_n_views = 700
views, views_level = sample_views(min_n_views, 450, azimuth_range, elev_range)
if vis:
save_vis(vis_path, views, views_level)
for n in range(len(views)):
R = views[n]['R']
t = views[n]['t']
mat_view = np.eye(4, dtype=np.float32)
mat_view[:3, :3] = R
mat_view[:3, 3] = t.squeeze()
# print(model)
img, xys = render_rgb(model, control_points, mat_view, shape=shape)
cv2.imwrite(rgb_path + '/' + str(n) + '.jpg', img)
with open(label_path + '/' + str(n) + '.txt', 'w') as f:
for i in xys:
f.write(str(i))
f.write(' ')
if __name__ == "__main__":
import fire
fire.Fire()