d-li14 / octconv.pytorch

PyTorch implementation of Octave Convolution with pre-trained Oct-ResNet and Oct-MobileNet models

octconv.pytorch's Issues

Could you add a example of resnet18?

resnet18 is quite different from resnet50.

BatchNorm before activation vs BatchNorm after activation

Thanks for your implementation of the Octave Conv paper.

I have a remark/question about the Conv_BN_ACT module.
As BN after ACT sometimes makes more sense, following the PyTorch example I made a small OctaveCNN (each using 6 convs total) for the CIFAR10 dataset. Using PReLU activations, CrossEntropy loss, AmsGrad optimizer and alpha=0.25.
After some experimentation with using BatchNorm before or after the activation I found the following results:

Network description	# epochs	accuracy (%)	training loss	test loss
Conv_BN_ACT	15	78.46	0.7093	0.6362
	30	82.20	0.4613	0.5456
Conv_ACT_BN	15	82.84	0.3917	0.5260
	30	84.18	0.1614	0.6036

I observe that Conv_ACT_BN has a tendency to overfit more as its training loss is noticeably lower than testing loss when compared to those of Conv_BN_ACT. However, Conv_ACT_BN does have a much higher accuracy.

Have you looked at this before? Is this the reason why you choose to include Conv_BN_ACT and not Conv_ACT_BN?

about the pre-trained model of the OctResNet-101

Thank you very much for your work.
When will the pre-trained model of the OctResNet-101 be released?

About loading pre-training errors

First of all, thank you for your code, which has benefited me a lot, but the same error always occurs when loading the pre-training model. The error code is as follows
Error(s) in loading state_dict for OctResNet:Missing key(s) in state_dict:
Can you help me with it, or can you provide me with a full version of the code? I am grateful! My email is [email protected]

about the pre-trained model

Thx for your great job! Where can i get the pretrained model on imagenet or other dataset?

Difference between your implementation and paper.

Hi,
when apply depthwise seperate conv, you directly abandon low-to-high and high-to-low branches. But in the paper, author compress high-to-high branch and preserve communication between high and low branch.

Training code

Are you planning on releasing the training code?

Also, did you try to implement the ResNet BasicBlock with OctConv?

I'm trying to do it in my own implementation, but it is tricky due to the lack of down sampling on the first layer.

Computation time around 3 times longer with OctConv

Thanks for your implementation of OctConv!

I see that the OctConv paper reports reduced GFLOPs usage as well as reduced computation time per image. However, running your implementation on ImageNet dataset, it seems the computation time per image actually increases by a multiple of 3! This happens both during training and during validation. I am comparing PyTorch's ResNet50 vs your OctResNet-50 with default alpha parameters.

Initialisation of layers and weights not available

I ran through the code and i couldnt find any snippet for initialisation of layers. do the layers need initialisation ?

and when "pretrain" parameter is set to TRUE, then nothing happens. Do we need to download the weights from somewhere ?

There is no main script to run the network ?

I don't see a main run script to run the training and evaluation for benchamarking. Seems you have create a module for octave convolution to plug into a custom written main script for benchmarking.

Good work , will you provide more pretrained models?

octave conv cannot support input at any size

If the network is downsampled n times, the input size must be 2^a (a >= n), other sizes are not supported, such as 600x1000.

redundant x_h in oct_resnet ?

Hi,
Is this the redundant line?

Could you please provide the training recipe file?

Thanks for your implementation of this work. I‘m using your code to reproduce the result you reported and I'm wondering if I could refer to your training settings. Could you please kindly provide your training config files if possible?

about plot

Thanks to the author for open source, as a newbie, I would like to ask how your four curves are drawn on a picture?thanks！

hi,Why the accuracy of original resnet26 is higher than octave-resnet26 used in cifar10 and cifar100?

cifar100 original_resnet26 77% oct_resnet26 65%
my training script is like this:
https://github.com/uoguelph-mlrg/Cutout/blob/master/train.py
could you shared your training script with us??
thank you.

Hi, why not adopting pre-actice version?

Why only return high frequency features (x_h)

def forward(self, x):
    x = self.conv1(x)
    x = self.bn1(x)
    x = self.relu(x)
    x = self.maxpool(x)

    x_h, x_l = self.layer1(x)
    x_h, x_l = self.layer2((x_h,x_l))
    x_h, x_l = self.layer3((x_h,x_l))
    x_h, x_l = self.layer4((x_h,x_l))
    x = self.avgpool(x_h)
    x = x.view(x.size(0), -1)
    x = self.fc(x)

    return x

train coed

first of all thanks for your implementation.
I want to train from scratch octconv.
So could you share train code?

Why is original resnet50 faster than octave-resnet50?

Why is original resnet50 faster than octave-resnet50?
In inference on gtx-1080, original resnet50 is about 70 fps and octave-resnet50 is around 42 fps.

basic_block in oct_resnet

Hi,
the BasicBlock here is the BasicBlock in pytorch restnet, right ? or something else ?

There are implementation with UNet?