def get_sample_region(self, gt, strides, num_points_per, gt_xs, gt_ys, radius=1):
num_gts = gt.shape[0]
K = len(gt_xs)
gt = gt[None].expand(K, num_gts, 4)
center_x = (gt[..., 0] + gt[..., 2]) / 2
center_y = (gt[..., 1] + gt[..., 3]) / 2
center_gt = gt.new_zeros(gt.shape)
# no gt
if center_x[..., 0].sum() == 0:
return gt_xs.new_zeros(gt_xs.shape, dtype='uint8')
beg = 0
for level, n_p in enumerate(num_points_per):
end = beg + n_p
stride = strides[level] * radius
xmin = center_x[beg:end] - stride
ymin = center_y[beg:end] - stride
xmax = center_x[beg:end] + stride
ymax = center_y[beg:end] + stride
# limit sample region in gt
center_gt[beg:end, :, 0] = jt.ternary(xmin > gt[beg:end, :, 0], xmin, gt[beg:end, :, 0])
center_gt[beg:end, :, 1] = jt.ternary(ymin > gt[beg:end, :, 1], ymin, gt[beg:end, :, 1])
center_gt[beg:end, :, 2] = jt.ternary(xmax > gt[beg:end, :, 2], gt[beg:end, :, 2], xmax)
center_gt[beg:end, :, 3] = jt.ternary(ymax > gt[beg:end, :, 3], gt[beg:end, :, 3], ymax)
beg = end
left = gt_xs[:, None] - center_gt[..., 0]
right = center_gt[..., 2] - gt_xs[:, None]
top = gt_ys[:, None] - center_gt[..., 1]
bottom = center_gt[..., 3] - gt_ys[:, None]
center_bbox = jt.stack((left, top, right, bottom), -1)
inside_gt_bbox_mask = center_bbox.min(-1)[0] > 0
return inside_gt_bbox_mask
def test_with_np(self):
np.random.seed(0)
a = np.random.rand(5,10).astype("float32")
b = np.random.rand(5,10).astype("float32")
ja = jt.array(a)
jb = jt.array(b)
jc = jt.ternary(ja>jb, ja, jb)
assert (jc.data==np.maximum(a,b)).all(), f"\n{jc.data}\n{np.maximum(a,b)}\n{a}\n{b}"
jda, jdb = jt.grad(jc, [ja, jb])
assert (jda.data==(a>b)*1).all()
assert (jdb.data==1-(a>b)).all()
def cross_entropy_loss(output, target, ignore_index=None):
if len(output.shape) == 4:
c_dim = output.shape[1]
output = output.transpose((0, 2, 3, 1))
output = output.reshape((-1, c_dim))
if ignore_index is not None:
target = jt.ternary(target == ignore_index,
jt.array(-1).broadcast(target), target)
mask = jt.logical_and(target >= 0, target < output.shape[1])
target = target.reshape((-1, ))
target = target.broadcast(output, [1])
target = target.index(1) == target
output = output - output.max([1], keepdims=True)
loss = output.exp().sum(1).log()
loss = loss - (output * target).sum(1)
if ignore_index is None:
return loss.mean()
else:
return loss.sum() / jt.maximum(mask.int().sum(), 1)
def leaky_relu(x, scale):
return jt.ternary(x > 0, x, x * scale)
def sign(x):
one = jt.ones(x.shape)
x = jt.ternary(x > 0, one, x)
return jt.ternary(x < 0, -one, x)
def leaky_relu(x, scale=0.01): return jt.ternary(x>0, x, x*scale) def relu6(x): return jt.minimum(jt.maximum(x, 0), 6)
def leaky_relu(x, scale): return jt.ternary(x>0, x, x*scale) #TODO dims is 4 will cause slowly execution def cross_entropy_loss(output, target, ignore_index=None):
def sign(x):
x = jt.ternary(x > 0, 1, x)
x = jt.ternary(x < 0, -1, x)
return x