marlin-data 8.0.0

Marlin Learn | Framework Data Adapter

marlin_data | the data adapter

© Rahul Tandon, RS Aqua 2024

About

marlin_data allows access to RSA's acoustic signature database. The module requests the RSA acoustic dataset required in order to run a machine learning (ML). Current version of marlin_data provides a predefined dataset for proof of concept and tutorial purposes. Future versions will allow for more specific datasets as well as definig training and validation datasets separately.

NOTE | Numerical data in marlin_data is defined using Python's numpy and pandas library.

Dependencies

import numpy as np import pandas as pd import requests, json import logging import dotenv import os, sys from dataclasses import dataclass import random

Installation

from marlin_data import *

Quick Start

Accessible Data

Datafeed Instance

Each iteration over the Marlin data feed will provide once instance of the datafeed and simulation data / snapshots. Frequency time series in numpy and a pandas dataframe is available along with descrictive metadata. Feed data is an iterable class which can be looped over allowing for an incremental datafeed.

Feed Instance -> frequency_ts_np

The frequency time series for this data snapshot ( NumPy Array )

    for feed_instance in data_feed: frequency_np_array = feed_instance['frequency_ts_np']

Feed Instance -> frequency_ts_pd

The frequency time series for this data snapshot (pandas DataFrame)

    for feed_instance in data_feed: frequency_data_frame = feed_instance['frequency_ts_pd']

Feed Instance -> meta_data

Metadata for the associated snapshot. Metadata is a dictionary with the following defined fields:

• snapshot_id : unique id for the snapshot
• data_frame_start : timestamp for the start of the snapshot data frame
• data_frame_end : timestamp for the end of the snapshot data frame
• listener_location : geo location of the listening device for snapshot
• location_name : human friendly name of listening location
• frame_delta_t : delta t for the snapshot (s)
• sample_rate : frequency recorder sample rate

Signature Data

SignatureData dataclass is accessible from MarlinData -> signature_data. Key values are stored in MarlinData -> signature_index.

    @dataclass
    class SignatureData:
            frequency_ts_np : np.array
            frequency_ts_pd : pd.DataFrame
            meta_data : {}

Metadata for the associated snapshot. Metadata is a dictionary with the following defined fields:

• snapshot_id : unique id for the snapshot
• data_frame_start : timestamp for the start of the snapshot data frame
• data_frame_end : timestamp for the end of the snapshot data frame
• listener_location : geo location of the listening device for snapshot
• location_name : human friendly name of listening location
• frame_delta_t : delta t for the snapshot (s)
• sample_rate : frequency recorder sample rate

Simulation Data

SimulationData dataclass is accessible from MarlinData -> simulation_data. Key values are stored in MarlinData -> simulation_index.

    @dataclass
    class SimulationData:
        frequency_ts_np : np.array
        frequency_ts_pd : pd.DataFrame
        meta_data : {}
        snapshot : bool = True

Metadata for the associated snapshot. Metadata is a dictionary with the following defined fields:

• snapshot_id : unique id for the snapshot
• data_frame_start : timestamp for the start of the snapshot data frame
• data_frame_end : timestamp for the end of the snapshot data frame
• listener_location : geo location of the listening device for snapshot
• location_name : human friendly name of listening location
• frame_delta_t : delta t for the snapshot (s)
• sample_rate : frequency recorder sample rate

Downloading data

Instantiate a MarlinData class.

    ` marlin_data = MarlinData(load_args={})`

Download signature data from RSA signature database.

    `marlin_data.download_signatures(load_args={})`

Download simulation / ML run data required for datafeed.

    `marlin_data.download_simulation_snapshots(load_args={})`

load_args:

• limit : maximum number of downloads (Require in init())
• signature_ids : vector of signature ids
• ss_ids : vector of snapshot ids◊
• location : vector of locations

Connecting to datafeed

Create the Marlin datafeed. This data feed can be iterated over in order to simulate a data feed into a model.

    `data_feed = MalrinDataStreamer()`

Connect the downloaded data to the datafeed instance.

    `data_feed.init_data(marlin_data.simulation_data, marlin_data.simulation_index) `

Iterate over the datafeed.

    for data_inst in data_feed:
            print (data_inst)

`