Closed DDXDB closed 5 months ago
For that you can simply refer to: https://github.com/qweasdd/manga-colorization-v2. It is the parent project.
You don't need GPU, when you will run the command python inference.py -p "path to file or folder"
, it defaults to CPU. If you do want to use GPU: python inference.py -p "path to file or folder" --gpu=True
I mean, do we need CUDA? I'm using the ARC A770, which is not a CUDA GPU but supports AI.a CUDA GPU but supports AI.
I think I did the parent project, but it doesn't seem to start:
Traceback (most recent call last): File "D:\Program Files\manga-colorization-v2-master\inference.py", line 5, in <module> import numpy as np ModuleNotFoundError: No module named 'numpy'
If you want to use GPU, then you need CUDA for this. Otherwise just run on CPU.
For the error, download this file.
And run pip install -r requirements.txt
You could also clone the kaggle notebook, run each cell one by one, till/before the APIS section. Then upload your folder to the kaggle notebook, create any new cell and add this code:
def run_batch_colorizer(path, use_cached):
class Configuration:
def __init__(self):
self.path = path
self.generator = 'networks/generator.zip'
self.extractor = 'networks/extractor.pth'
self.gpu = True
self.denoiser = True
self.denoiser_sigma =25
self.size = 576
self.use_cached = use_cached
args = Configuration()
if args.gpu:
device = 'cuda'
else:
device = 'cpu'
colorizer = MangaColorizator(device, args.generator, args.extractor)
colorization_path = os.path.join('colored')
os.makedirs(colorization_path, exist_ok=True)
batch_colorize_images(colorization_path, colorizer, args)
Run the cell. Then make a new cell with (enter your folder path):
run_batch_colorizer(the_uploaded_folder_path, True)
After running this, you should get all the colored images and can download the 'colored' folder.
I have a lot of downloaded comic collections saved as zip, is this project available for local comics?
and,need CUDA GPU?