Our great sponsors
-
InfluxDB
Power Real-Time Data Analytics at Scale. Get real-time insights from all types of time series data with InfluxDB. Ingest, query, and analyze billions of data points in real-time with unbounded cardinality.
I am very glad that my work aroused interest. Thank you all! I will try to answer the questions. How did i create this? I collected a dataset of about 600 images of dark art sketches, processed them so that they would be suitable for training in the StyleGAN2-ada (resized to 1024x1024, edited something a little, made sure that all images have 3 channels). Mainly used photoshop, also Duplicate Photos Fixer Pro to find duplicates, and also I highly recommend Derek's dataset-tools for preparing datasets. Then the dataset was archived, uploaded to GoogleDrive and added to Google Colab for training. I had to subscribe to the Colab Pro because the free version could not start training due to lack of memory. A pro subscription costs $10 per month and provides advantages in capacity and uptime. More about working with Google Colab can be found here. I'm a beginner myself, so no secret techniques have been applied. In fact, everything is as in the video tutorial. Training took place about 30 hours, partly at Tesla P100 and partly on a more powerful Tesla V100. I have not written anywhere an article about this project because I do not speak English very well. And there is not much to write about, everything is simple. In the future I will probably post the .pkl file to the public.
It seems like this even would make for a cool website, like pix2pix
Related posts
- Any work on Style transfer using Stable Diffusion based on image-mask pairs similar to Pix2Pix
- Hello nooby programmer here but i do 3d art and i have few questions
- Predict Fourier spectrum
- Things I Have Drawn is a site in which the things kids draw are real
- Is it possible to remove a sticker from a photo?