ristoranterist/fastflow

fastflow's Issues

Custom datasets

Hello, I made an error with custom datasets, how can I solve it?

How to run train.py

TypeError: init() missing 2 required positional arguments: 'ac' and 'config'

Loss in fastflow

FastFlow/utils.py

Line 8 in 86396fe

return (squared_error - jacob, squared_error, jacob)

In the paper in the introduction section they describe in the contributions:

"We propose a 2D normalizing flow denoted as FastFlow
for anomaly detection and localization with fully convolutional networks and two-dimensional loss function to
effectively model global and local distribution"

Is there a reason why the loss isn't computed pixelwise prior to mean reduction. Right now you mean reduce both the Z.^2 and J terms separately prior to computing the difference. From the phrasing of the paper, i would have expected this to be written as:

def calc_loss(z,j):
    return torch.sum(0.5*z**-j,(1,2,3))

Obviously I'm guessing since the paper doesn't describe specifics, but just thought I'd check how you arrived at the current form of the loss?

datasets？why？

test implementation

Hi, could I use test implementation? I have a problem with save_path. Did you use 500 epochs for training phase based on the paper?

RuntimeError: CUDA out of memory.

When I run it, it raises RuntimeError: CUDA out of memory. Tried to allocate 74.00 MiB (GPU 0; 11.91 GiB total capacity; 10.79 GiB already allocated; 12.94 MiB free; 11.21 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
I want to know how large the memory of gpu is needed.

Why do you use the GLOW design?

Hello. Is good to know that there's people trying to reproduce this work. I'm also working on it.

Empty auroc images

Hello, after executing the code main.py, multiple files get saved in the results folder. But all the auroc images are empty..why is that? I would be very grateful for someone's help