Filter-Response-Normalization-PyTorch

Unofficial Pytorch implementation of the paper Filter Response Normalization(https://arxiv.org/abs/1911.09737).

Example

Using convert_model function

import torch
from torch import nn
import numpy as np
import torchvision
from filter_response_normalization import convert_model
m = torchvision.models.resnet18(True)
m = nn.DataParallel(m)
m = convert_model(m)
x = np.zeros((3, 3, 384, 576), dtype="f")
x = torch.from_numpy(x)
y = m(x)
print(y.size()) # Output:torch.Size([3, 1000])

Issue

• Change input shape of TLU
  In paper implement, threshold shape of ThresholdedLinearUnit is (1, channels, 1, 1)
  But in this implement, ThresholdedLinearUnit threshold is scolor. So if you do not use the convert_model function, use ThresholdedLinearUnitFixND.
  It was solved by qubvel contribution. 🎉

Distributed under MIT License (See LICENSE)