Source code for meerkat.datasets.rfw

import os
import subprocess
from multiprocessing.sharedctypes import Value
from typing import Dict

import numpy as np
import pandas as pd

import meerkat as mk

from ..abstract import DatasetBuilder
from ..info import DatasetInfo
from ..registry import datasets
from ..utils import download_url, extract


@datasets.register()
class rfw(DatasetBuilder):

    VERSIONS = ["main"]

    GROUPS = ["Caucasian", "African", "Asian", "Indian"]

    info = DatasetInfo(
        name="fer",
        full_name="Racial Faces in-the-Wild",
        # flake8: noqa
        description="Racial Faces in-the-Wild (RFW) is a testing database for studying racial bias in face recognition. Four testing subsets, namely Caucasian, Asian, Indian and African, are constructed, and each contains about 3000 individuals with 6000 image pairs for face verification. They can be used to fairly evaluate and compare the recognition ability of the algorithm on different races.",
        # flake8: noqa
        homepage="http://www.whdeng.cn/RFW/testing.html",
        tags=["image", "facial recognition", "algorithmic bias"],
    )

    def build(self):

        dfs = []
        for group in self.GROUPS:
            df = pd.read_csv(
                os.path.join(self.dataset_dir, f"test/txts/{group}/{group}_images.txt"),
                delimiter="\t",
                names=["filename", "count"],
            )

            df["ethnicity"] = group.lower()
            df["identity"] = df["filename"].str.rsplit("_", n=1).str[0]
            df["image_id"] = df["filename"].str.rsplit(".", n=1).str[0]
            df["image_path"] = df.apply(
                lambda x: f"test/data/{group}/{x['identity']}/{x['filename']}", axis=1
            )
            df.drop(columns=["filename", "count"])
            dfs.append(df)
        df = pd.concat(dfs)
        dp = mk.DataPanel.from_pandas(df)
        dp["image"] = mk.ImageColumn.from_filepaths(
            dp["image_path"], base_dir=self.dataset_dir
        )
        return dp[["image_id", "identity", "ethnicity", "image"]]

        return None

    def download(self):
        raise ValueError(
            "To download the RFW dataset, you must request access following the "
            "instructions at http://www.whdeng.cn/RFW/testing.html."
            "Once you've been granted access and downloaded the data, move it "
            f"to the directory {self.dataset_dir} and extract it."
        )