Numerical encoding of the FairFace labels

twitter_saliency_filter_analysis--08.py

twitter_saliency_eval_dir = f'{img_dir}//Twitter_saliency'
if not os.path.exists(twitter_saliency_eval_dir):
  print(f'No outputs directory: {twitter_saliency_eval_dir} found ...')
  execute_in_shell([f'mkdir {twitter_saliency_eval_dir}'])
  print(f'Created outputs directory: {twitter_saliency_eval_dir}')
  
labels_encoder_file = f'{twitter_saliency_eval_dir}/labels_encoder.json'
if os.path.exists(labels_encoder_file):
  with open(labels_encoder_file) as f:
    labels_encoder = json.loads(f.read())
  print(labels_encoder)
  print(f'Loaded labels encoder data from: {labels_encoder_file} ...')  
else:
  print(f'No saved labels encoder data: {labels_encoder_file} ...')
  labels_encoder = {}
  for i, race in enumerate(sorted(list(set(img_labels['race'].values)))):
    labels_encoder.update({race: i})
  print(labels_encoder)
  with open(labels_encoder_file, 'w+') as f:
    json.dump(labels_encoder, f)
  print(f'Saved labels encoder data to: {labels_encoder_file} ...')
  
def encoded_labels(input_label, labels_encoder):
  return labels_encoder[input_label]
def decoded_labels(input_label, labels_encoder):
  return list(labels_encoder.keys())[list(labels_encoder.values()).index(input_label)]