Data preparation for the Inference Pipeline

This module contains functions for data ingestion and preprocessing during the inference phase.

Data Ingestion

• mov_to_mp3_audio: Extracts audio from a video file and saves it as an mp3 file.

from emotion_detective.data.inference.data_ingestion import mov_to_mp3_audio

# Example usage
mov_to_mp3_audio('input_video.mp4')

emotion_detective.data.inference.data_ingestion.mov_to_mp3_audio(input_file: str) → str

Extracts audio from a video file and saves it as an mp3 file.

Parameters:

input_file (str) – Path to the input video file.

Returns:

Path to the saved mp3 file.

Return type:

str

Raises:

• ValueError – If the input file does not have a .mp4 extension.

• Exception – Any exceptions encountered during the conversion process.

Author: Kacper Janczyk

Troubleshooting

• Problem: FileNotFoundError when the input video file is not found. Solution: Ensure the input file path is correct and the file exists.

  try:
      mov_to_mp3_audio('input_video.mp4')
  except FileNotFoundError:
      print("The specified video file was not found. Please check the file path.")
  

Data Preprocessing

• transcribe_translate: Transcribes and translates audio files.

from emotion_detective.data.inference.data_preprocessing import transcribe_translate

# Example usage
transcribe_translate('audio_file.mp3')

emotion_detective.data.inference.data_preprocessing.dataset_loader(df: DataFrame, text_column: str, batch_size=1) → DataLoader

Creates a PyTorch DataLoader for a given pandas DataFrame.

Parameters: - df (pandas.DataFrame): The DataFrame containing the dataset. - text_column (str): The name of the column in df containing text data. - batch_size (int): The batch size to be used for loading the data.

Returns: - torch.utils.data.DataLoader: A DataLoader object configured with

the provided DataFrame and batch size.

Author: Kacper Janczyk

emotion_detective.data.inference.data_preprocessing.transcribe_translate(path: str, language: str = 'en') → DataFrame

Transcribes and translates audio files.

Parameters: - path (str): The path to the audio file for transcription. - language (str, optional): The language of the audio and the target

language for translation. Default is ‘en’ (English).

Returns: - pandas.DataFrame: A DataFrame containing transcribed and translated

sentences along with their corresponding start and end times.

Author: Kacper Janczyk

Troubleshooting

• Problem: FileNotFoundError when the audio file is not found. Solution: Ensure the input file path is correct and the file exists.

  try:
      transcribe_translate('audio_file.mp3')
  except FileNotFoundError:
      print("The specified audio file was not found. Please check the file path.")
  

• Problem: UnsupportedFormatError when the audio file format is not supported. Solution: Convert the audio file to a supported format such as Mp3.

  try:
      transcribe_translate('audio_file.wav')
  except UnsupportedFormatError:
      print("The audio file format is not supported. Please convert it to Mp3 and try again.")
  

• dataset_loader: Creates a PyTorch DataLoader for a given DataFrame.

from emotion_detective.data.inference.data_preprocessing import dataset_loader

# Example usage
loader = dataset_loader(df, 'text_column', batch_size=32)

Troubleshooting

• Problem: KeyError when the specified text column is not found in the DataFrame. Solution: Verify that the DataFrame contains the specified text column.

  try:
      loader = dataset_loader(df, 'text_column', batch_size=32)
  except KeyError:
      print("The specified text column was not found in the DataFrame. Please check the column name.")
  

• Problem: ValueError when the batch size is invalid. Solution: Ensure that the batch size is a positive integer.

  try:
      loader = dataset_loader(df, 'text_column', batch_size=32)
  except ValueError:
      print("Invalid batch size. Please ensure the batch size is a positive integer.")