# it was originally in 
# /home/junhyeok/datasets/turn-taking-datasets/EASYCOM_TT/target_perframe/_event_preprocess.ipynb
import os
import numpy as np

label_dir = 'class5'
label_files = os.listdir(label_dir)

print(label_files)

events = {}
for label_file in label_files:
  label_path = os.path.join(label_dir, label_file)
  
  labels = np.load(label_path)
  
  target_speaker_labels = labels[:, 1]
  
  
  print(label_file, len(target_speaker_labels))
  event = []
  for i in range(1, len(target_speaker_labels)):
    # state transition from 0 to 1 is the start of the event
    if (target_speaker_labels[i-1] == 0 and target_speaker_labels[i] == 1):
      event.append([i, 'target speaker speaking start'])
    # state transition from 1 to 0 is the end of the event  
    if (target_speaker_labels[i-1] == 1 and target_speaker_labels[i] == 0):
      event.append([i-1, 'target speaker speaking end'])
  
  # save as numpy file
  event = np.array(event)
  print(f'label_file: {label_file}, event shape: {event.shape}')
  np.save(os.path.join('events_remove_bc', label_file), event)