Run inference on images using trained model

run_inference.py

mport PIL
import glob
import os
from tqdm.notebook import tqdm

render_factor = 40

if os.path.exists('imagepaths.txt'):
  os.remove('imagepaths.txt')

!rm -Rf seinfeld_inference/high_res/
!mkdir seinfeld_inference/high_res/

def write_to_txt(dest):
  file = open("imagepaths.txt", "a") 
  write = "file '" + dest.strip() + "'" + "\n"
  file.write(write) 
  file.close() 

files = sorted(glob.glob('seinfeld_inference/images/*.*g'), key = lambda x: int(os.path.basename(x).split('.')[0]))
files = files[300:]

for i in tqdm(range(1000)):

  # file = random.choice(files)
  file = files[i]
  dest = 'seinfeld_inference/high_res/'+os.path.basename(file)

  # scale to square
  new_path = scale_to_square(PIL.Image.open(file), render_factor*16, dest = dest.split('.')[0]+'_square.jpg')

  # run inference
  run_inference_images(new_path, dest)

  # unsquare
  dest = unsquare(PIL.Image.open(dest), PIL.Image.open(file), dest = dest.split('.')[0]+'_unsquared.jpg')

  # write to txt
  write_to_txt(dest)

  # increase brightness
  adjust_brightness(PIL.Image.open(dest),factor = 1.75, dest=dest)