Roman
chore: include reshape in FHE execution
1974e22 unverified
raw
history blame
9.6 kB
"Filter definitions, with pre-processing, post-processing and compilation methods."
import numpy as np
import torch
from torch import nn
from common import AVAILABLE_FILTERS, INPUT_SHAPE
from concrete.numpy.compilation.compiler import Compiler
from concrete.ml.common.utils import generate_proxy_function
from concrete.ml.torch.numpy_module import NumpyModule
class TorchIdentity(nn.Module):
"""Torch identity model."""
def forward(self, x):
"""Identity forward pass.
Args:
x (torch.Tensor): The input image.
Returns:
x (torch.Tensor): The input image.
"""
return x
class TorchInverted(nn.Module):
"""Torch inverted model."""
def forward(self, x):
"""Forward pass for inverting an image's colors.
Args:
x (torch.Tensor): The input image.
Returns:
torch.Tensor: The (color) inverted image.
"""
return 255 - x
class TorchRotate(nn.Module):
"""Torch rotated model."""
def forward(self, x):
"""Forward pass for rotating an image.
Args:
x (torch.Tensor): The input image.
Returns:
torch.Tensor: The rotated image.
"""
return x.transpose(0, 1)
class TorchConv(nn.Module):
"""Torch model with a single convolution operator."""
def __init__(self, kernel, n_in_channels=3, n_out_channels=3, groups=1, threshold=None):
"""Initialize the filter.
Args:
kernel (np.ndarray): The convolution kernel to consider.
"""
super().__init__()
self.kernel = torch.tensor(kernel, dtype=torch.int64)
self.n_out_channels = n_out_channels
self.n_in_channels = n_in_channels
self.groups = groups
self.threshold = threshold
def forward(self, x):
"""Forward pass with a single convolution using a 1D or 2D kernel.
Args:
x (torch.Tensor): The input image.
Returns:
torch.Tensor: The filtered image.
"""
# Define the convolution parameters
stride = 1
kernel_shape = self.kernel.shape
# Ensure the kernel has a proper shape
# If the kernel has a 1D shape, a (1, 1) kernel is used for each in_channels
if len(kernel_shape) == 1:
kernel = self.kernel.reshape(
self.n_out_channels,
self.n_in_channels // self.groups,
1,
1,
)
# Else, if the kernel has a 2D shape, a single (Kw, Kh) kernel is used on all in_channels
elif len(kernel_shape) == 2:
kernel = self.kernel.expand(
self.n_out_channels,
self.n_in_channels // self.groups,
kernel_shape[0],
kernel_shape[1],
)
else:
raise ValueError(
"Wrong kernel shape, only 1D or 2D kernels are accepted. Got kernel of shape "
f"{kernel_shape}"
)
# Reshape the image. This is done because Torch convolutions and Numpy arrays (for PIL
# display) don't follow the same shape conventions. More precisely, x is of shape
# (Width, Height, Channels) while the conv2d operator requires an input of shape
# (Batch, Channels, Height, Width)
x = x.transpose(2, 0).unsqueeze(axis=0)
# Apply the convolution
x = nn.functional.conv2d(x, kernel, stride=stride, groups=self.groups)
# Reshape the output back to the original shape (Width, Height, Channels)
x = x.transpose(1, 3).reshape((x.shape[2], x.shape[3], self.n_out_channels))
# Subtract a given threshold if given
if self.threshold is not None:
x -= self.threshold
return x
class Filter:
"""Filter class used in the app."""
def __init__(self, filter_name):
"""Initializing the filter class using a given filter.
Most filters can be found at https://en.wikipedia.org/wiki/Kernel_(image_processing).
Args:
filter_name (str): The filter to consider.
"""
assert filter_name in AVAILABLE_FILTERS, (
f"Unsupported image filter or transformation. Expected one of {*AVAILABLE_FILTERS,}, "
f"but got {filter_name}",
)
# Define attributes associated to the filter
self.filter_name = filter_name
self.onnx_model = None
self.fhe_circuit = None
self.divide = None
self.repeat_out_channels = False
# Instantiate the torch module associated to the given filter name
if filter_name == "identity":
self.torch_model = TorchIdentity()
elif filter_name == "inverted":
self.torch_model = TorchInverted()
elif filter_name == "rotate":
self.torch_model = TorchRotate()
elif filter_name == "black and white":
# Define the grayscale weights (RGB order)
# These weights were used in PAL and NTSC video systems and can be found at
# https://en.wikipedia.org/wiki/Grayscale
# There are initially supposed to be float weights (0.299, 0.587, 0.114), with
# 0.299 + 0.587 + 0.114 = 1
# However, since FHE computations require weights to be integers, we first multiply
# these by a factor of 1000. The output image's values are then divided by 1000 in
# post-processing in order to retrieve the correct result
kernel = [299, 587, 114]
self.torch_model = TorchConv(kernel, n_out_channels=1, groups=1)
# Define the value used when for dividing the output values in post-processing
self.divide = 1000
# Indicate that the out_channels will need to be repeated, as Gradio requires all
# images to have a RGB format, even for grayscaled ones
self.repeat_out_channels = True
elif filter_name == "blur":
kernel = np.ones((3, 3))
self.torch_model = TorchConv(kernel, n_out_channels=3, groups=3)
# Define the value used when for dividing the output values in post-processing
self.divide = 9
elif filter_name == "sharpen":
kernel = [
[0, -1, 0],
[-1, 5, -1],
[0, -1, 0],
]
self.torch_model = TorchConv(kernel, n_out_channels=3, groups=3)
elif filter_name == "ridge detection":
kernel = [
[-1, -1, -1],
[-1, 9, -1],
[-1, -1, -1],
]
# Additionally to the convolution operator, the filter will subtract a given threshold
# value to the result in order to better display the ridges
self.torch_model = TorchConv(kernel, n_out_channels=1, groups=1, threshold=900)
# Indicate that the out_channels will need to be repeated, as Gradio requires all
# images to have a RGB format, even for grayscaled ones. Ridge detection images are
# ususally displayed as such
self.repeat_out_channels = True
def compile(self):
"""Compile the filter on a representative inputset."""
# Generate a random representative set of images used for compilation, following Torch's
# shape format (batch, in_channels, image_height, image_width) for each samples
# This version's compiler only handles tuples of 1-batch array as inputset, meaning we need
# to define the inputset as a Tuple[np.ndarray[shape=(1, 3, H, W)]]
np.random.seed(42)
# inputset = tuple(
# np.random.randint(0, 256, size=((1, 3) + INPUT_SHAPE), dtype=np.int64) for _ in range(100)
# )
inputset = tuple(
np.random.randint(0, 256, size=(INPUT_SHAPE + (3, )), dtype=np.int64) for _ in range(100)
)
# Convert the Torch module to a Numpy module
numpy_module = NumpyModule(
self.torch_model,
dummy_input=torch.from_numpy(inputset[0]),
)
# Get the proxy function and parameter mappings used for initializing the compiler
# This is done in order to be able to provide any modules with arbitrary numbers of
# encrypted arguments to Concrete Numpy's compiler
numpy_filter_proxy, parameters_mapping = generate_proxy_function(
numpy_module.numpy_forward,
["inputs"]
)
# Compile the filter and retrieve its FHE circuit
compiler = Compiler(
numpy_filter_proxy,
{parameters_mapping["inputs"]: "encrypted"},
)
self.fhe_circuit = compiler.compile(inputset)
return self.fhe_circuit
def post_processing(self, output_image):
"""Apply post-processing to the encrypted output images.
Args:
input_image (np.ndarray): The decrypted image to post-process.
Returns:
input_image (np.ndarray): The post-processed image.
"""
# Divide all values if needed
if self.divide is not None:
output_image //= self.divide
# Clip the image's values to proper RGB standards as filters don't handle such constraints
output_image = output_image.clip(0, 255)
# Gradio requires all images to follow a RGB format
if self.repeat_out_channels:
output_image = output_image.repeat(3, axis=2)
return output_image