config_inference.yaml

#######################################################################################
############ CONFIG FILE FOR HUMAN SPEECH DETECTIONS AND DATA ANONYMIZATION ###########
#######################################################################################

###################
# Path parameters #
###################

# Path to the data to be anonymised
PATH_INPUT_DATA: "./assets/demo_data/data_to_anonymised"

# Path to the result files, namely json files containing the start and end of detections
# One json file is created per analysed file
PATH_JSON_DETECTIONS: "./assets/demo_data/detections/json/ecoVAD"

# Path to store the anonymised data (i.e. audiofiles where human speech has clearly been removed)
PATH_ANONYMIZED_DATA: "./assets/demo_data/anonymised_data"

##############################################
# Algorithm to use in anonymisation.py script #
##############################################

# VAD algorithm to use for detecting human speech
# In its current state this includes "ecovad", "pyannote" or "webrtcvad"
CHOSEN_VAD: "ecovad"
# For the chosen VAD model, make sure you tune the parameters below

##########################
# ecoVAD hyperparameters #
##########################

# Path to ecoVAD weights, the default weights are stored in ./assets/model_weights/
# Note that you can also use your own model weights
ECOVAD_WEIGHTS_PATH: "./assets/model_weights/ecoVAD_ckpt.pt"

# Confidence threshold.
THRESHOLD: 0.7

#############################
# WebrtcVAD hyperparameters #
#############################
# For more details about webrtcvad parameters refer to the github repository
# https://github.com/wiseman/py-webrtcvad

# A frame must be either 10, 20, or 30 ms in duration
FRAME_LENGTH: 30

# aggressiveness mode, which is an integer between 0 and 3. 
# 0 is the least aggressive about filtering out non-speech, 3 is the most aggressive
AGGRESSIVENESS: 1

#######################
# Hardware parameters #
#######################

# Whether ecoVAD and pyannote should be used on GPU
USE_GPU: False


##################################################################################
# Parameters for extracting some detections from the extract_detection.py script #
##################################################################################

# The created file will be a CSV containing all the information from the detection json files
# Useful if one wants to visualise the result 
PATH_PARSED_JSON: "./assets/demo_data/detections/parsed_json/parsed_json.csv"

# Path where the detections should be stored
PATH_SAMPLE_DETECTIONS: "./assets/demo_data/sample_detections"

# Number of samples that should be taken
NUMBER_OF_SAMPLES: 5

# Random seed
RANDOM_SEED: 42