import os
import pandas as pd
import matplotlib.pyplot as plt
from tqdm import tqdm
import ast
# === Load CSV ===
input_path = "maslow_image_level_output_deduplicated.csv"
df = pd.read_csv(input_path)
# === Maslow level mappings ===
level_name_map = {
"level_1": "physiological_needs",
"level_2": "safety_needs",
"level_3": "love_and_belonging",
"level_4": "esteem_needs",
"level_5": "self-actualization",
"level_6": "self-Transcendence"
}
maslow_levels = list(level_name_map.values())
# === Colours for plotting ===
maslow_colors = {
"physiological_needs": "#e6194B",
"safety_needs": "#f58231",
"love_and_belonging": "#ffe119",
"esteem_needs": "#3cb44b",
"self-actualization": "#911eb4",
"self-Transcendence": "#8c564b"
}
# === Define your two groups ===
group_1 = [
"Connection_images", "Creativity", "Education", "freedom", "happiness",
"human_instinct", "instinct", "nature", "play", "safety"
]
group_2 = [
"technology_addiction", "doomscrolling", "social_media_addiction", "conformity",
"overstimulation", "standardisation", "isolation", "danger", "artifical", "sadness", "disconnection"
]
# === Function to process one group ===
def process_group(group_list, group_label):
percent_data = []
for category in tqdm(group_list, desc=f"Processing {group_label}"):
subset = df[df["category"] == category]
image_level_weights = []
for _, row in subset.iterrows():
image_weights = {}
total_objects = 0
for raw_level in level_name_map:
try:
obj_list = ast.literal_eval(row[raw_level]) if pd.notna(row[raw_level]) else []
obj_count = len(obj_list)
except:
obj_count = 0
image_weights[raw_level] = obj_count
total_objects += obj_count
if total_objects > 0:
for raw_level in image_weights:
image_weights[raw_level] /= total_objects
image_level_weights.append(image_weights)
avg_weights = {friendly: 0 for friendly in maslow_levels}
for weights in image_level_weights:
for raw_level, friendly_name in level_name_map.items():
avg_weights[friendly_name] += weights[raw_level]
num_images = len(image_level_weights)
if num_images > 0:
for k in avg_weights:
avg_weights[k] = round(100 * avg_weights[k] / num_images, 1)
avg_weights["category"] = category
percent_data.append(avg_weights)
# Create and save percent df
percent_df = pd.DataFrame(percent_data).set_index("category").sort_index()
percent_df.to_csv(f"{group_label}_per_image_percent.csv")
# Global normalisation across each Maslow level
level_totals = percent_df[maslow_levels].sum()
normalised_df = percent_df[maslow_levels].div(level_totals) * 100
normalised_df = normalised_df.round(1)
normalised_df.to_csv(f"{group_label}_double_normalised.csv")
# Plot
ax = normalised_df.plot(
kind="bar",
figsize=(14, 7),
edgecolor="black",
color=[maslow_colors[lvl] for lvl in maslow_levels]
)
plt.title(f"{group_label} Category Distribution per Maslow Level (Double Normalised)")
plt.xlabel("Category")
plt.ylabel("Percent of Total for Each Maslow Level")
plt.xticks(rotation=45, ha="right")
plt.legend(title="Maslow Level")
plt.tight_layout()
plt.savefig(f"{group_label}_maslow_chart.png")
plt.close()
# === Run both groups ===
process_group(group_1, "group1")
process_group(group_2, "group2")