Secure your code as it's written. Use Snyk Code to scan source code in minutes - no build needed - and fix issues immediately.
axes = self._normalize_axes(img, axes)
axes_net = self.config.axes
_permute_axes = self._make_permute_axes(axes, axes_net)
x = _permute_axes(img) # x has axes_net semantics
channel = axes_dict(axes_net)['C']
self.config.n_channel_in == x.shape[channel] or _raise(ValueError())
axes_net_div_by = self._axes_div_by(axes_net)
grid = tuple(self.config.grid)
len(grid) == len(axes_net)-1 or _raise(ValueError())
grid_dict = dict(zip(axes_net.replace('C',''),grid))
normalizer = self._check_normalizer_resizer(normalizer, None)[0]
resizer = StarDistPadAndCropResizer(grid=grid_dict)
x = normalizer.before(x, axes_net)
x = resizer.before(x, axes_net, axes_net_div_by)
def predict_direct(tile):
sh = list(tile.shape); sh[channel] = 1; dummy = np.empty(sh,np.float32)
prob, dist = self.keras_model.predict([tile[np.newaxis],dummy[np.newaxis]], **predict_kwargs)
return prob[0], dist[0]
if np.prod(n_tiles) > 1:
tiling_axes = axes_net.replace('C','') # axes eligible for tiling
x_tiling_axis = tuple(axes_dict(axes_net)[a] for a in tiling_axes) # numerical axis ids for x
axes_net_tile_overlaps = self._axes_tile_overlap(axes_net)
# hack: permute tiling axis in the same way as img -> x was permuted
n_tiles = _permute_axes(np.empty(n_tiles,np.bool)).shape
(all(n_tiles[i] == 1 for i in range(x.ndim) if i not in x_tiling_axis) or