Select Page

DA Weighted Code 

import os
import pandas as pd
import numpy as np

# Load your dataset
df = pd.read_csv("All_object_DA_categorie.csv")

# Define which columns are object-related
object_columns = df.columns[3:]

# Step 1: Count how many images are in each category
category_counts = df['category'].value_counts().to_dict()

# Step 2: Count total number of individual objects per image (splitting on commas)
def count_objects(row):
    total = 0
    for col in object_columns:
        val = row[col]
        if pd.notna(val) and isinstance(val, str):
            items = [item.strip() for item in val.split(',') if item.strip()]
            total += len(items)
    return total

df['num_objects'] = df.apply(count_objects, axis=1)

# Step 3: Compute object_value (no extra columns)
df['object_value'] = [
    (1 / category_counts[cat]) * (1 / num) if num > 0 else np.nan
    for cat, num in zip(df['category'], df['num_objects'])
]

# Optional: remove the 'num_objects' column if you don't want to keep it
df = df.drop(columns=['num_objects'])

# Save or inspect
df.to_csv("All_object_DA_categorie_with_each_object_weighted.csv", index=False)
print(df[['image', 'category', 'object_value']].head())


←→