def points2bbox(self, pts, y_first=True):
"""点集转换成包围框.
:param pts: the input points sets (fields), each points
set (fields) is represented as 2n scalar.
:param y_first: if y_first=True, the point set is represented as
[y1, x1, y2, x2 ... yn, xn], otherwise the point set is
represented as [x1, y1, x2, y2 ... xn, yn].
:return: each points set is converting to a bbox [x1, y1, x2, y2].
"""
pts_reshape = L.reshape(pts, (pts.shape[0], -1, 2, pts.shape[2], pts.shape[3]))
pts_y = pts_reshape[:, :, 0, :, :] if y_first else pts_reshape[:, :, 1, :, :]
pts_x = pts_reshape[:, :, 1, :, :] if y_first else pts_reshape[:, :, 0, :, :]
if self.transform_method == 'minmax':
# bbox_left = pts_x.min(dim=1, keepdim=True)[0]
# bbox_right = pts_x.max(dim=1, keepdim=True)[0]
# bbox_up = pts_y.min(dim=1, keepdim=True)[0]
# bbox_bottom = pts_y.max(dim=1, keepdim=True)[0]
# bbox = torch.cat([bbox_left, bbox_up, bbox_right, bbox_bottom],
# dim=1)
pass
elif self.transform_method == 'partial_minmax':
# pts_y = pts_y[:, :4, ...]
# pts_x = pts_x[:, :4, ...]
# bbox_left = pts_x.min(dim=1, keepdim=True)[0]
# bbox_right = pts_x.max(dim=1, keepdim=True)[0]
# bbox_up = pts_y.min(dim=1, keepdim=True)[0]
# bbox_bottom = pts_y.max(dim=1, keepdim=True)[0]
# bbox = torch.cat([bbox_left, bbox_up, bbox_right, bbox_bottom],
# dim=1)
pass
elif self.transform_method == 'moment':
pts_y_mean = L.reduce_mean(pts_y, dim=1, keep_dim=True)
pts_x_mean = L.reduce_mean(pts_x, dim=1, keep_dim=True)
pts_y_std = paddle.std(pts_y - pts_y_mean, axis=1, keepdim=True)
pts_x_std = paddle.std(pts_x - pts_x_mean, axis=1, keepdim=True)
moment_transfer = (self.moment_transfer * self.moment_mul) + (
self.moment_transfer.detach() * (1 - self.moment_mul))
moment_width_transfer = moment_transfer[0]
moment_height_transfer = moment_transfer[1]
half_width = pts_x_std * L.exp(moment_width_transfer)
half_height = pts_y_std * L.exp(moment_height_transfer)
bbox = L.concat([
pts_x_mean - half_width, pts_y_mean - half_height,
pts_x_mean + half_width, pts_y_mean + half_height
], axis=1)
else:
raise NotImplementedError
return bbox
def static(self):
with paddle.static.program_guard(paddle.static.Program()):
x = paddle.data('X', self.shape, self.dtype)
out = paddle.std(x, self.axis, self.unbiased, self.keepdim)
exe = paddle.static.Executor(self.place)
res = exe.run(feed={'X': self.x}, fetch_list=[out])
return res[0]
def get_logic_from_model(name):
x, _ = load_sample_audio_text()
model = Wav2Vec2ForCTC(name, pretrained=True)
model.eval()
with paddle.no_grad():
logits = model(x)
return float(paddle.mean(logits)), float(paddle.std(logits))
def dynamic(self, axis=None, keepdim=False, unbiased=True):
with fluid.dygraph.guard(self._place):
data = fluid.dygraph.to_variable(self._input)
out = paddle.std(input=data,
axis=axis,
keepdim=keepdim,
unbiased=unbiased)
return out.numpy()
def static(self, axis=None, keepdim=False, unbiased=True):
prog = fluid.Program()
with fluid.program_guard(prog):
data = fluid.data(name="data",
dtype=self._dtype,
shape=[None, 3, 4, 5])
out = prog.current_block().create_var(dtype=self._dtype,
shape=[2, 3, 4, 5])
paddle.std(input=data,
axis=axis,
keepdim=keepdim,
unbiased=unbiased,
out=out)
exe = fluid.Executor(self._place)
return exe.run(feed={"data": self._input},
program=prog,
fetch_list=[out])[0]
def test_alias(self):
paddle.disable_static()
x = paddle.to_tensor(np.array([10, 12], 'float32'))
out1 = paddle.std(x).numpy()
out2 = paddle.tensor.std(x).numpy()
out3 = paddle.tensor.stat.std(x).numpy()
self.assertTrue(np.allclose(out1, out2))
self.assertTrue(np.allclose(out1, out3))
paddle.enable_static()
def forward(self, x): # (b, c, h, w)
b, c, h, w = x.shape
x = paddle.reshape(x, [b * self.groups, -1, h, w])
xn = x * self.avg_pool(x)
xn = paddle.sum(xn, axis=1, keepdim=True)
t = paddle.reshape(xn, [b * self.groups, -1])
t = t - paddle.mean(t, axis=1, keepdim=True)
std = paddle.std(t, axis=1, keepdim=True) + 1e-5
t = t / std
t = paddle.reshape(t, [b, self.groups, h, w])
t = t * self.weight + self.bias
t = paddle.reshape(t, (b * self.groups, 1, h, w))
x = x * self.sig(t)
x = paddle.reshape(x, (b, c, h, w))
return x
def convertDistribution(self, x):
mean, std = paddle.mean(x), paddle.std(x)
y = (x - mean) * 0.2 / std
return y.astype('float32')
def dygraph(self):
paddle.disable_static()
x = paddle.to_tensor(self.x)
out = paddle.std(x, self.axis, self.unbiased, self.keepdim)
paddle.enable_static()
return out.numpy()