def permutate_and_take_min(h: np.array, A: np.array, B: np.array) -> np.array:
hashes = h.repeat(A.shape[0]).reshape(h.shape[0], A.shape[0])
hashes = (A * hashes + B) % _mersenne_prime
# minhashes = np.empty(hashes.shape[1])
# for i in range(hashes.shape[1]):
# minhashes[i] = hashes[:, i].min()
minhashes = [hashes[:, i].min() for i in range(hashes.shape[1])]
return minhashes
def to_tensor(self,
arr: np.array,
normalize: bool = False,
repeat_channels: int = None) -> torch.Tensor:
"""
Transform an array to a torch Tensor of type Float64 and shape [C, W, H]
"""
if normalize:
arr = arr / arr.max()
else:
arr = arr * 1.0
arr = torch.tensor(arr, device=self._device)
if len(arr.shape) <= 2:
arr = arr.unsqueeze(0)
if repeat_channels:
arr = arr.repeat(repeat_channels, 1, 1)
return arr
