/Upscale

Upscale

Argument scale:super resolution scale,choice from 2 or 4,default is "4",

img_dir:path to folder containing pictures

img_type:image format，default is "png"

output_dir:the result path

device_index: gpu id, default is "0"

model_type: cnn or gan, defalut is "cnn"

python3.6 main.py --scale 4 --img_dir=path_to_img --output_dir=path_to_result --model_type=cnn

the tutorial of upscaling images in your GPU server

Step1

pull docker image dokcer pull registry.cn-hangzhou.aliyuncs.com/rotoscope/upscale:v1.1 before that you need to have this permission. We have done at 192.168.2.150

Step2

run doccker docker run -dit -v /data/upscale:/data --name upscale registry.cn-hangzhou.aliyuncs.com/rotoscope/upscale:v1.1 the dir /data/upscale is your host path. We have done at 192.168.2.150

Step3

prepare your images, make a dir in /data/upscale at your host server. such as apple, ad apple/result.

Then, run docker exec -it upscale python3.6 main.py --scale 4 --img_dir=/data/apple--output_dir=/data/apple/result --model_type=cnn --img_type=jpg

  • scale: super resolution scale,2 or 4
  • img_type: your image ext, jpg or png
  • img_dir: your image(s) path
  • model_type: cnn or gan
  • output_dir: the result of your image(s) path