def id_to_rgb(cls, id):
"""
Cover float[-2**20~2**20] to float32[0.~1.] rgb
Parameters
----------
id : int, float or numpy
float[-2**20~2**20].
Returns
-------
(3,)float32
RGB for Blender.
"""
if isinstance(id, (float, int)):
id = np.array(id)
rgb = np.zeros(id.shape + (3, ), dtype=np.float64)
rgb[..., 0] = id < 0
absf = np.abs(id)
int_part = absf // 1
# rgb[..., 1] = 1 - 1 / (int_part + 1)
poww = np.int32(np.log2(int_part + 1, dtype=np.float32)) + 1
denominator = (2**poww).round()
rgb[..., 1] = ((int_part - denominator // 2 + 1) * 2 + 1) / denominator
rgb[..., 2] = absf % 1
return rgb
def id_to_rgb(f):
if isinstance(f, (float, int)):
f = np.array(f)
rgb = np.zeros(f.shape + (3, ), dtype=np.float64)
rgb[..., 0] = f < 0
absf = np.abs(f)
int_part = absf // 1
rgb[..., 1] = 1 - 1 / (int_part + 1)
rgb[..., 2] = absf % 1
return rgb
def forward(self, feats, xyens):
logName = 'cyc_r: %s, out_cyc_r: %s' % (self.cyc_r, self.out_cyc_r
or 'N')
logTag = timegap(self.log_freq, logName)
if logTag:
logDic = dicto()
if self.layerNorma:
feats = layerNormaFun(feats)
tensorType = feats.type()
shape = self.shape = feats.shape
feats = feats.view(-1, *shape[-2:])
xyens = xyens.view(-1, 3)
pgts = xyens[..., [1, 0]].cpu().numpy()
existMask = xyens[..., -1].type(tensorType)
# with timeit(logName):
masks = map2(getMaskOfPgt, [
dict(pgt=pgt,
cycle_tmpl=self.cycle_tmpl,
shape=shape,
out_cyc_r=self.out_cyc_r) for pgt in pgts
])
# masks = mapmp(self.getMask, pgts, pool=4)
masks = np.array(masks)
masks = th.from_numpy(np.uint8(masks)).type(tensorType).cuda()
loss = 0
#(lambda a,b,t=1:e**(t*a)/(e**(t*a)+e**(t*b)))(2,1,5)
def softmaxFgBgOld(fg, bg, t=1):
fge = th.exp(fg * t)
bge = th.exp(bg * t)
prob = fge / (fge + bge + eps)
return prob
def softmaxFgBgSubMax(fg, bg, t=1):
fg = fg * t
bg = bg * t
maxx = max(float(fg.max()), float(bg.max()))
fge = th.exp(fg - maxx)
bge = th.exp(bg - maxx)
prob = fge / (fge + bge + eps)
return prob
def softmaxFgBgSub(fg, bg, t=1):
diff = bg - fg
toExp = t * diff
if (toExp > 80).sum() and timegap(1, 'toExp'):
from boxx import prettyFrameLocation, pred
print(prettyFrameLocation())
pred - "toExp.max() is %.2f > 80, diff.max() is %.2f" % (
toExp.max(), diff.max())
prob = 1 / (1 + th.exp(toExp))
return prob
softmaxFgBg = softmaxFgBgSubMax
def CE(fg, bg):
prob = softmaxFgBg(fg, bg)
avgLosses = -th.log(prob + eps)
return avgLosses
if 'avg' in self.poolings:
bgAvgPool = (feats * masks[..., 0, :, :]).sum(-1).sum(-1) / (
masks[..., 0, :, :].sum(-1).sum(-1) + eps)
fgAvgPool = (feats * masks[..., 1, :, :]).sum(-1).sum(-1) / (
masks[..., 1, :, :].sum(-1).sum(-1) + eps)
avgProbs = softmaxFgBg(fgAvgPool, bgAvgPool, self.temper)
avgLosses = -th.log(avgProbs + eps)
indexMask = existMask * (avgProbs < self.probMargin).type(
tensorType) if self.probMargin else existMask
avgLoss = (avgLosses * indexMask).sum() / (indexMask.sum() + eps)
loss += avgLoss
if logTag:
logDic.avgLoss = float(avgLoss)
logDic.avgProb = float(avgProbs.mean())
if 'max' in self.poolings:
bgMaxPool = (feats * masks[..., 0, :, :]).max(-1)[0].max(-1)[0]
fgMaxPool = (feats * masks[..., 1, :, :]).max(-1)[0].max(-1)[0]
maxProbs = softmaxFgBg(fgMaxPool, bgMaxPool, self.temper)
maxLosses = -th.log(maxProbs + eps)
indexMask = existMask * (maxProbs < self.probMargin).type(
tensorType) if self.probMargin else existMask
maxLoss = (maxLosses * indexMask).sum() / (indexMask.sum() + eps)
loss += maxLoss
if logTag:
logDic.maxLoss = float(maxLoss)
logDic.maxProb = float(maxProbs.mean())
if logTag:
print("%s | %s" % (logName, ', '.join(
map(lambda kv: "%s: %.3f" % kv, logDic.items()))))
# print(Markdown([{k:strnum(v) for k,v in logDic.items()}]))
# g()
return loss
hw = Vector((9, 9))
channl = 1
batch = 1
multi_batch = False
multi_batch = True
if multi_batch:
gpun = 1
batchPerGpu = 1
channl = 1
if cloud:
batchPerGpu = 32
channl = 17
gpun = th.cuda.device_count()
batch = batchPerGpu * gpun
cyc_r = intround((25) / np.array([353, 257]).mean() * hw.mean())
pgt = hw // 2
pshift = (hw * .13).intround()
pshift = Vector([0, 0])
pre = pgt + pshift
coreShape = Vector((
cyc_r * 2,
cyc_r * 2,
))
feat = np.zeros([batch, channl] + list(hw))
raw_cycle_tmpl = getWeightCore(300, seta=.25)
# raw_cycle_tmpl = raw_cycle_tmpl > raw_cycle_tmpl[150,0]
core = resize(raw_cycle_tmpl, coreShape)
msegt = feat.copy()
colormap[a][0] + (colormap[b][0] - colormap[a][0]) * f,
colormap[a][1] + (colormap[b][1] - colormap[a][1]) * f,
colormap[a][2] + (colormap[b][2] - colormap[a][2]) * f,
]
def interpolate_or_clip(colormap, x):
if x < 0.0:
return [0.0, 0.0, 0.0]
elif x > 1.0:
return [1.0, 1.0, 1.0]
else:
return interpolate(colormap, x)
turbo_colormap_data_np = np.array(turbo_colormap_data)
def heatmap_to_pseudo_color(heatmap):
x = heatmap
x = x.clip(0, 1)
a = (x * 255).astype(int)
b = (a + 1).clip(max=255)
f = x * 255.0 - a
pseudo_color = (
turbo_colormap_data_np[a] +
(turbo_colormap_data_np[b] - turbo_colormap_data_np[a]) * f[..., None])
pseudo_color[heatmap < 0.0] = 0.0
pseudo_color[heatmap > 1.0] = 1.0
cube_size = 1
for inst_id in range(1, 3):
location = [random.random() * 4 - 2 for _ in range(3)]
rotation = [random.random() * 2 * pi for _ in range(3)]
# add cube
bpy.ops.mesh.primitive_cube_add(size=cube_size,
location=location,
rotation=rotation)
obj = bpy.context.active_object
# set each instance a unique inst_id, which is used to generate instance annotation.
obj["inst_id"] = inst_id
# result["ycb_meta"] is 6d pose GT
result = bpycv.render_data()
meta = result["ycb_meta"]
img = result["image"]
# cube mesh xyz
cube_xyz_s = "111 110 101 011 100 010 001 000"
cube_xyz = np.array([[-cube_size / 2, cube_size / 2][int(s)]
for s in cube_xyz_s if s in "01"]).reshape(-1, 3)
cube_xyz = None
vis = vis_poses_by_meta(img, meta, cube_xyz)
cv2.imwrite("demo-vis_6d_pose.jpg", vis)
if __name__ == "__main__":
pass
def f(imgn):
imread(imgn)
if __name__ == '__main__':
imgns = imgns[:]
# with timeit():
# images = mapmp(f, imgns, pool=16)
# with timeit():
# images = mapmp(f, imgns, pool=8)
# with timeit():
# images = mapmt(f, imgns, pool=16)
with timeit():
images = mapmt(f, imgns, pool=8)
# with timeit():
# images = map2(f, imgns, )
images = np.array((images))
attrp = pathjoin(st_gan_dataset, 'attribute_train.npy')
imagep = p / pathjoin(st_gan_dataset, 'image_train.npy')
np.save(attrp, attribute)
np.save(imagep, images)
# os.link(attrp, attrp.replace('train.npy', 'test.npy'))
# os.link(imagep, imagep.replace('train.npy', 'test.npy'))
if __name__ == "__main__":
pass
# add cube
bpy.ops.mesh.primitive_cube_add(size=cube_size,
location=location,
rotation=rotation)
obj = bpy.context.active_object
# set each instance a unique inst_id, which is used to generate instance annotation.
obj["inst_id"] = inst_id
result = bpycv.render_data()
# result["ycb_meta"] is 6d pose GT
meta = result["ycb_meta"]
img = result["image"]
# cube mesh xyz
cube_xyz = np.array([
[1.0, 1.0, 1.0],
[1.0, 1.0, -1.0],
[1.0, -1.0, 1.0],
[-1.0, 1.0, 1.0],
[1.0, -1.0, -1.0],
[-1.0, 1.0, -1.0],
[-1.0, -1.0, 1.0],
[-1.0, -1.0, -1.0],
])
cube_xyz *= cube_size / 2
vis = vis_poses_by_ycb_meta(img, meta, cube_xyz)
cv2.imwrite("demo-vis_6d_pose.jpg", vis)
def adjustBrightnessPil(pil, target):
new = adjustBrightnessInHsv(np.array(pil), adjustBrightness)
newpil = Image.fromarray(new)
return newpil, target
