Image Metadata Extraction

Raw Data → Ingestion → Validation → Cleaning → Transformation → Output
              │            │            │            │
              ▼            ▼            ▼            ▼
          Log errors   Flag issues  Fix/Remove   Enrich data

pip install pandas numpy openpyxl pyarrow

import pandas as pd
import numpy as np
from typing import Optional
import logging

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)


class DataCleaner:
    """Automated data cleaning pipeline for Image Metadata Extraction."""

    def __init__(self, df: pd.DataFrame):
        self.df = df.copy()
        self.report = {"original_rows": len(df), "steps": []}

    def remove_duplicates(self, subset: Optional[list] = None) -> "DataCleaner":
        before = len(self.df)
        self.df = self.df.drop_duplicates(subset=subset)
        removed = before - len(self.df)
        self._log("remove_duplicates", removed)
        return self

    def handle_missing(
        self, strategy: str = "drop", fill_value=None, columns: Optional[list] = None
    ) -> "DataCleaner":
        cols = columns or self.df.columns.tolist()
        before = self.df[cols].isnull().sum().sum()

        if strategy == "drop":
            self.df = self.df.dropna(subset=cols)
        elif strategy == "fill":
            self.df[cols] = self.df[cols].fillna(fill_value)
        elif strategy == "median":
            for col in cols:
                if pd.api.types.is_numeric_dtype(self.df[col]):
                    self.df[col] = self.df[col].fillna(self.df[col].median())
        elif strategy == "mode":
            for col in cols:
                self.df[col] = self.df[col].fillna(self.df[col].mode().iloc[0])

        self._log(f"handle_missing ({strategy})", int(before))
        return self

    def remove_outliers(self, columns: list, method: str = "iqr", factor: float = 1.5) -> "DataCleaner":
        before = len(self.df)
        for col in columns:
            if method == "iqr":
                Q1 = self.df[col].quantile(0.25)
                Q3 = self.df[col].quantile(0.75)
                IQR = Q3 - Q1
                mask = (self.df[col] >= Q1 - factor * IQR) & (self.df[col] <= Q3 + factor * IQR)
                self.df = self.df[mask]
            elif method == "zscore":
                z = np.abs((self.df[col] - self.df[col].mean()) / self.df[col].std())
                self.df = self.df[z < factor]
        self._log("remove_outliers", before - len(self.df))
        return self

    def standardize_text(self, columns: list) -> "DataCleaner":
        for col in columns:
            self.df[col] = (
                self.df[col]
                .str.strip()
                .str.lower()
                .str.replace(r"\s+", " ", regex=True)
            )
        self._log("standardize_text", 0)
        return self

    def get_report(self) -> dict:
        self.report["final_rows"] = len(self.df)
        self.report["rows_removed"] = self.report["original_rows"] - len(self.df)
        return self.report

    def _log(self, step: str, affected: int):
        logger.info(f"[{step}] affected {affected} rows")
        self.report["steps"].append({"step": step, "affected": affected})

def main():
    # Load raw data
    df = pd.read_csv("data/raw_dataset.csv")
    print(f"Loaded {len(df)} rows, {len(df.columns)} columns")

    # Run cleaning pipeline
    cleaner = DataCleaner(df)
    clean_df = (
        cleaner
        .remove_duplicates()
        .handle_missing(strategy="median", columns=["revenue", "quantity"])
        .handle_missing(strategy="mode", columns=["category"])
        .remove_outliers(columns=["revenue"], method="iqr")
        .standardize_text(columns=["name", "category"])
        .df
    )

    # Output
    report = cleaner.get_report()
    print(f"Cleaning complete: {report['original_rows']} → {report['final_rows']} rows")
    clean_df.to_parquet("data/clean_dataset.parquet", index=False)


if __name__ == "__main__":
    main()

def validate(df: pd.DataFrame) -> list[str]:
    issues = []
    if df.duplicated().any():
        issues.append(f"Found {df.duplicated().sum()} duplicates")
    for col in df.select_dtypes(include="number"):
        null_pct = df[col].isnull().mean() * 100
        if null_pct > 5:
            issues.append(f"{col}: {null_pct:.1f}% null values")
    return issues