Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

refusing some images, but some not. #40

Open
HerXayah opened this issue Dec 3, 2021 · 14 comments
Open

refusing some images, but some not. #40

HerXayah opened this issue Dec 3, 2021 · 14 comments

Comments

@HerXayah
Copy link

HerXayah commented Dec 3, 2021

Everytime i try to use Step 4 e.g
python train.py --dataset lilly --epoch 101 --init_epoch 10
i get to the point of generating them etc but it always fails on the checkpoint.
I could not find how to generate one.
heres my log: https://www.toptal.com/developers/hastebin/raw/udidakecec

I would like to add there is nothing saved in the folder i specified nor anywhere else like you said in #36 on the last comment.

@TachibanaYoshino
Copy link
Owner

image
I think this runtime error has nothing to do with checkpoints. The model will automatically save the checkpoints only after each epoch is completed. Your error here appears in the process of data loading. Please check whether your data is missing, such as whether the style_smooth data exists.

@HerXayah
Copy link
Author

HerXayah commented Dec 3, 2021

image
They both exist and are popupated.
The files itself are jpgs in both.

@HerXayah
Copy link
Author

HerXayah commented Dec 3, 2021

Ok so Update.
It seems like it doesnt like .jpeg. it only takes .jpg although the twitter downloader i used gave me .jpeg. weird.
Wouldnt it be easier to convert every image to png or jpg, regardless of what extension they had?

EDIT:
So sometimes it straight up refuses images. I converted the tenet trailer into png`s, then ran them over the edge smooth and same error.

EDIT2:
Now its just straight up not liking .jpg file too.
Ill send you the images im trying with here. Theyre converted from mp4 through ffmpeg via ffmpeg -i input.mp4 -vf fps=5 style/elon%06d.jpg
https://drive.google.com/file/d/11nOdlG27C7Z-nwn-SuJsZ2n5o_BwnizY/view?usp=sharing

@HerXayah HerXayah changed the title [*] Failed to find a checkpoint refusing some images, but some not. Dec 3, 2021
@TachibanaYoshino
Copy link
Owner

if path.split('.')[-1] not in ['jpg', 'jpeg', 'png', 'gif']:

You can see through this code that it supports jpg and jpeg. I don't know how your data is placed. If it is convenient, I can remotely control and check your code and data through teamViewer, and send me the teamViewer account and password via email.

@HerXayah
Copy link
Author

HerXayah commented Dec 4, 2021

if path.split('.')[-1] not in ['jpg', 'jpeg', 'png', 'gif']:

You can see through this code that it supports jpg and jpeg. I don't know how your data is placed. If it is convenient, I can remotely control and check your code and data through teamViewer, and send me the teamViewer account and password via email.

We can do that sure.
Im just git cloning this repo though as far as code goes.
using a conda env with python 3.6 with my installed packages here
A friend of mine also suggested that its my dataset that is wrong. Yesterday i tried to make a new dataset using only images i got from reddit via rightclick -> download.
image
Here you can see my dataset,. which in theory, should have worked.
Also is there a way to direct message on Github or something for the teamviewer id :D

@TachibanaYoshino
Copy link
Owner

I think it is very likely that the ImageGenerator class did not read the path of your image, or the path is set incorrectly somewhere.

@HerXayah
Copy link
Author

HerXayah commented Dec 4, 2021

I think it is very likely that the ImageGenerator class did not read the path of your image, or the path is set incorrectly somewhere.

Why does it work with some of them and when i add more it doesnt? Thats whats confusing me the most

@TachibanaYoshino
Copy link
Owner

I think it is very likely that the ImageGenerator class did not read the path of your image, or the path is set incorrectly somewhere.

Why does it work with some of them and when i add more it doesnt? Thats whats confusing me the most

That's so sad. I don't know your specific situation.

@HerXayah
Copy link
Author

HerXayah commented Dec 4, 2021

I can try and Zip one where it works and then add stuff until it not works and make a seperate folder.
Im not using a different version from the one thats in here so we should we able to find out whats the cause in this way

@HerXayah
Copy link
Author

HerXayah commented Dec 4, 2021

It seems i found something.
If your smooth directory has more than 10 files it refuses to work. (in this case atleast? i had more files once i think)
Ive included a Zip with showcase below.
i tried with the last image in the file list and the first one and it always worked, but the moment i add more file so its 11 files, it will stop working and error me
https://drive.google.com/file/d/11nOdlG27C7Z-nwn-SuJsZ2n5o_BwnizY/view?usp=sharing

@HerXayah
Copy link
Author

HerXayah commented Dec 7, 2021

any update?

@HerXayah
Copy link
Author

any idea?

@zobnechopin
Copy link

any idea?

@PrincessAkira @TachibanaYoshino

I found something interesting about this and run the training process successful.For the training process, you need prepare three part dataset: style, smooth and real ,the error shows checkout point error, but indeed the actually error is real data not load.
So, the solution is download the whole dataset on Github and put test, read data under your train dataset dir.
But after this I meet another error that shows anime dataset 'tuple out of index', I checked all my dataset and source code, and I found nothing...But, after I checked dataset, it works.

@HerXayah
Copy link
Author

any idea?

@PrincessAkira @TachibanaYoshino

I found something interesting about this and run the training process successful.For the training process, you need prepare three part dataset: style, smooth and real ,the error shows checkout point error, but indeed the actually error is real data not load. So, the solution is download the whole dataset on Github and put test, read data under your train dataset dir. But after this I meet another error that shows anime dataset 'tuple out of index', I checked all my dataset and source code, and I found nothing...But, after I checked dataset, it works.

I can try that later today!
thanks

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants