binitdox/manga-colorizer

Bring mangas to life!

Hello there 👋

• 👯 I’m looking to collaborate on any deep learning research
• ⚡ Fun fact: Best way to learn about something is to make an interesting project

manga-colorizer's Issues

KAGGLE not working

when I run the kaggle notebook it says

RuntimeError Traceback (most recent call last)
Cell In[16], line 10
7 import ssl
8 import threading
---> 10 from colorizator import MangaColorizator
11 from inference import colorize_images as batch_colorize_images

File /kaggle/working/Manga-Colorizer/Backend/colorizator.py:1
----> 1 import torch
2 from torchvision.transforms import ToTensor
3 import numpy as np

File /opt/conda/lib/python3.10/site-packages/torch/init.py:229
227 if USE_GLOBAL_DEPS:
228 _load_global_deps()
--> 229 from torch._C import * # noqa: F403
231 # Appease the type checker; ordinarily this binding is inserted by the
232 # torch._C module initialization code in C
233 if TYPE_CHECKING:

RuntimeError: THPDtypeType.tp_dict == nullptr INTERNAL ASSERT FAILED at "/usr/local/src/pytorch/torch/csrc/Dtype.cpp":135, please report a bug to PyTorch.

any idea how to fix it please?

Help?

Anyone to contact you, I need help setting this up as I'm having major issues.

Discord: travis#6256

extension is not working

it keeps saying ''Error!" on https://chapmanganelo.com/ and i don't know server url

SSL Certificate outdated

SSL Certificate outdated

Can be applied to local downloaded comics

I have a lot of downloaded comic collections saved as zip, is this project available for local comics?

and，need CUDA GPU？

Had some trouble installing the self-hosting version, but now it works and it is awesome!

Following the "Server Usage Instructions (SELF HOSTING, OLD)"
at the step "pip install -r requirements.txt" I got the error:
*** Please install the scikit-image package (instead of skimage) ***
I solved this by:
pip install scikit-image # this step was probably not needed since it is in the requirements
then removing the line for skimage from requirements.txt
then running again:
pip install -r requirements.txt
It warned that things were installed ~/.local/bin but it was not in my path, so I made sure it was in my path.

After this I ran into problems with the versions of Flask and Werkzeug (ImportError: cannot import name 'url_quote' from 'werkzeug.urls') and finally solved this by updating to the latest versions of both:
pip uninstall flask Flask_Cors Werkzeug
pip install flask Flask_Cors Werkzeug

Next I had an "ee key too small" error and regenerated the server key:
in Backend folder:
openssl req -nodes -newkey rsa:2048 -keyout server.key -out server.csr
(pressed Enter to use default/blank values for most fields; provided a new challenge password but haven't been asked for it yet)
then made a new self-signed server certificate:
openssl x509 -req -days 730 -in server.csr -signkey server.key -out server.crt

Next I had the problem that the version of PyTorch installed did not support my GPU, so I removed it
pip3 uninstall torch torchvision
and updated to the 11.8 version using a command generated by https://pytorch.org/get-started/locally/
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
(Pretty sure I could have left out torchaudio, but I was quick to paste the command and run it.)

Now it runs and it is awesome! I am very impressed with how quickly it colorizes a whole chapter and how good it looks!

I have not tested this idea, but I think all the version issues would be solved by removing version numbers from the lines in requirements.txt for scikit-image, flask, Flask_Cors, Werkzeug, torch, and torchvision and removing the whole line for skimage.