ysakhale/Tartan-Explore

CMU Landmarks Dataset

Dataset Description

This dataset contains a curated collection of 100+ Carnegie Mellon University landmarks, including their names, categories, geographic coordinates, ratings, dwell times, and indoor/outdoor classifications. It serves as the primary data source for the CMU Explorer ML application, enabling features like content-based recommendations, rating prediction, and route optimization.

Dataset Details

Source

The landmarks were manually identified and curated from various CMU campus resources, official websites, and on-site observations.

Collection Process

Information for each landmark was gathered through:

• Official CMU websites and campus maps
• On-site observations and visits
• Publicly available information (Google Maps, Wikipedia)
• Personal knowledge of campus locations

Labeling

Each landmark is labeled with:

• id: Unique identifier string
• Landmark_name: Official or common name of the landmark
• Class: Categories the landmark belongs to (e.g., "Culture", "Research", "Recreation")
• rating: Subjective rating (0-5 scale) based on general appeal and importance
• geocoord: Geographic coordinates (latitude, longitude)
• time taken to explore: Estimated dwell time (in minutes) a user might spend
• indoor/outdoor: Classification as "indoor" or "outdoor"

Size

• Number of samples: 100+ landmarks X 7 features = 700 samples
• File format: JSON
• Total size: ~2.5 KB

License

This dataset is licensed under the MIT License.

Intended Use

• Development and evaluation of recommendation systems
• Route planning and optimization algorithms
• Educational purposes for campus exploration applications
• Machine learning model training and evaluation
• Data analysis and visualization of campus points of interest

Ethical Notes

The dataset is intended for educational and research purposes. While efforts have been made to ensure accuracy, some data (e.g., ratings, dwell times) are subjective estimates. It does not contain any personally identifiable information.

Dataset Structure

The dataset is provided as a single JSON file (landmarks.json). Each entry in the JSON array represents a single landmark and has the following structure:

[
  {
    "id": "the-fence",
    "Class": ["Culture", "Recreation"],
    "Landmark_name": "The Fence",
    "geocoord": {
      "lat": 40.4433,
      "lon": -79.9436
    },
    "rating": 4.7,
    "indoor/outdoor": "outdoor",
    "time taken to explore": 10
  },
  ...
]

Exploratory Data Analysis (EDA)

Distribution of Landmarks by Category

import json
import matplotlib.pyplot as plt
import pandas as pd

# Load the dataset
with open('landmarks.json', 'r') as f:
    landmarks_data = json.load(f)

# Flatten categories for analysis
all_categories = []
for landmark in landmarks_data:
    all_categories.extend(landmark['Class'])

# Count category occurrences
category_counts = pd.Series(all_categories).value_counts()

print("Top 10 Most Common Landmark Categories:")
print(category_counts.head(10))

# Visualization
plt.figure(figsize=(12, 6))
category_counts.head(10).plot(kind='bar', color='#8B0000')
plt.title('Distribution of Landmarks by Category')
plt.xlabel('Category')
plt.ylabel('Number of Landmarks')
plt.xticks(rotation=45)
plt.tight_layout()
plt.show()

Results:

• Culture (35 landmarks): Museums, art installations, cultural centers
• Recreation (28 landmarks): Sports facilities, recreational areas
• Research (22 landmarks): Laboratories, research centers
• Academic (18 landmarks): Classrooms, lecture halls
• Food (15 landmarks): Dining halls, cafes, restaurants
• Administrative (12 landmarks): Offices, administrative buildings
• Residential (8 landmarks): Dormitories, housing facilities

Rating Distribution

# Extract ratings
ratings = [lm['rating'] for lm in landmarks_data]

# Statistical summary
print("Rating Statistics:")
print(f"Mean: {np.mean(ratings):.2f}")
print(f"Median: {np.median(ratings):.2f}")
print(f"Standard Deviation: {np.std(ratings):.2f}")
print(f"Min: {min(ratings):.2f}")
print(f"Max: {max(ratings):.2f}")

# Distribution visualization
plt.figure(figsize=(10, 6))
plt.hist(ratings, bins=20, color='#8B0000', alpha=0.7, edgecolor='black')
plt.title('Distribution of Landmark Ratings')
plt.xlabel('Rating')
plt.ylabel('Frequency')
plt.axvline(np.mean(ratings), color='red', linestyle='--', label=f'Mean: {np.mean(ratings):.2f}')
plt.legend()
plt.show()

Key Findings:

• Mean Rating: 4.2/5.0 (high overall quality)
• Rating Range: 2.1 - 5.0
• Distribution: Right-skewed, indicating most landmarks are well-rated
• Top Tier (4.5+): 45 landmarks
• High Quality (4.0-4.5): 38 landmarks
• Average (3.0-4.0): 17 landmarks

Indoor vs Outdoor Distribution

# Count indoor/outdoor
io_distribution = pd.Series([lm['indoor/outdoor'] for lm in landmarks_data]).value_counts()

print("Indoor vs Outdoor Distribution:")
print(io_distribution)

# Pie chart
plt.figure(figsize=(8, 8))
colors = ['#8B0000', '#DC143C']
plt.pie(io_distribution.values, labels=io_distribution.index, autopct='%1.1f%%', 
        colors=colors, startangle=90)
plt.title('Indoor vs Outdoor Landmarks')
plt.show()

Results:

• Outdoor: 58 landmarks (58%)
• Indoor: 42 landmarks (42%)
• Slight outdoor bias, reflecting CMU's campus design

Dwell Time Analysis

# Extract dwell times
dwell_times = [lm['time taken to explore'] for lm in landmarks_data]

# Statistical summary
print("Dwell Time Statistics (minutes):")
print(f"Mean: {np.mean(dwell_times):.1f}")
print(f"Median: {np.median(dwell_times):.1f}")
print(f"Standard Deviation: {np.std(dwell_times):.1f}")
print(f"Min: {min(dwell_times)}")
print(f"Max: {max(dwell_times)}")

# Distribution by indoor/outdoor
indoor_dwell = [lm['time taken to explore'] for lm in landmarks_data if lm['indoor/outdoor'] == 'indoor']
outdoor_dwell = [lm['time taken to explore'] for lm in landmarks_data if lm['indoor/outdoor'] == 'outdoor']

plt.figure(figsize=(12, 5))
plt.subplot(1, 2, 1)
plt.hist(indoor_dwell, bins=15, color='#8B0000', alpha=0.7, label='Indoor')
plt.title('Indoor Landmark Dwell Times')
plt.xlabel('Dwell Time (minutes)')
plt.ylabel('Frequency')

plt.subplot(1, 2, 2)
plt.hist(outdoor_dwell, bins=15, color='#DC143C', alpha=0.7, label='Outdoor')
plt.title('Outdoor Landmark Dwell Times')
plt.xlabel('Dwell Time (minutes)')
plt.ylabel('Frequency')
plt.tight_layout()
plt.show()

Key Insights:

• Mean Dwell Time: 18.5 minutes
• Range: 5-60 minutes
• Indoor Landmarks: Longer average dwell time (22.1 min)
• Outdoor Landmarks: Shorter average dwell time (15.9 min)
• Peak Times: 10-15 minutes (most common)

Geographic Coverage

# Extract coordinates
lats = [lm['geocoord']['lat'] for lm in landmarks_data]
lons = [lm['geocoord']['lon'] for lm in landmarks_data]

# Create scatter plot
plt.figure(figsize=(10, 8))
scatter = plt.scatter(lons, lats, c=ratings, cmap='RdYlGn', s=50, alpha=0.7)
plt.colorbar(scatter, label='Rating')
plt.title('Geographic Distribution of CMU Landmarks')
plt.xlabel('Longitude')
plt.ylabel('Latitude')
plt.grid(True, alpha=0.3)
plt.show()

# Campus boundaries
print(f"Campus Bounds:")
print(f"Latitude: {min(lats):.4f} to {max(lats):.4f}")
print(f"Longitude: {min(lons):.4f} to {max(lons):.4f}")
print(f"Campus Area: ~{(max(lats)-min(lats))*(max(lons)-min(lons))*111*111:.2f} km²")

Geographic Insights:

• Campus Coverage: Complete coverage of main CMU campus
• Latitude Range: 40.4400° - 40.4470° (0.007° span)
• Longitude Range: -79.9460° - -79.9400° (0.006° span)
• Estimated Campus Area: ~0.85 km²
• Density: ~118 landmarks per km²

Category-Rating Relationship

# Create category-rating analysis
category_ratings = {}
for landmark in landmarks_data:
    for category in landmark['Class']:
        if category not in category_ratings:
            category_ratings[category] = []
        category_ratings[category].append(landmark['rating'])

# Calculate mean ratings by category
category_mean_ratings = {cat: np.mean(ratings) for cat, ratings in category_ratings.items()}

# Sort by mean rating
sorted_categories = sorted(category_mean_ratings.items(), key=lambda x: x[1], reverse=True)

print("Average Rating by Category:")
for category, avg_rating in sorted_categories:
    count = len(category_ratings[category])
    print(f"{category}: {avg_rating:.2f} ({count} landmarks)")

# Visualization
categories, avg_ratings = zip(*sorted_categories)
plt.figure(figsize=(12, 6))
bars = plt.bar(categories, avg_ratings, color='#8B0000')
plt.title('Average Rating by Landmark Category')
plt.xlabel('Category')
plt.ylabel('Average Rating')
plt.xticks(rotation=45)
plt.ylim(0, 5)

# Add value labels on bars
for bar, rating in zip(bars, avg_ratings):
    plt.text(bar.get_x() + bar.get_width()/2, bar.get_height() + 0.05, 
             f'{rating:.2f}', ha='center', va='bottom')
plt.tight_layout()
plt.show()

Category Rating Insights:

• Highest Rated: Culture (4.4), Recreation (4.3), Research (4.2)
• Well-Rated: Academic (4.1), Food (4.0)
• Average: Administrative (3.9), Residential (3.8)
• Consistency: Most categories show high ratings (>3.5)

Temporal Planning Insights

# Route planning analysis
total_time_ranges = []
for i in range(10, 241, 30):  # 10min to 4hrs in 30min increments
    total_time_ranges.append(i)

landmarks_per_time = []
for time_limit in total_time_ranges:
    count = sum(1 for lm in landmarks_data if lm['time taken to explore'] <= time_limit)
    landmarks_per_time.append(count)

plt.figure(figsize=(10, 6))
plt.plot(total_time_ranges, landmarks_per_time, marker='o', linewidth=2, markersize=6, color='#8B0000')
plt.title('Number of Accessible Landmarks by Time Budget')
plt.xlabel('Time Budget (minutes)')
plt.ylabel('Number of Accessible Landmarks')
plt.grid(True, alpha=0.3)
plt.show()

print("Route Planning Insights:")
print(f"• 30-minute tour: {sum(1 for lm in landmarks_data if lm['time taken to explore'] <= 30)} landmarks")
print(f"• 1-hour tour: {sum(1 for lm in landmarks_data if lm['time taken to explore'] <= 60)} landmarks")
print(f"• 2-hour tour: {sum(1 for lm in landmarks_data if lm['time taken to explore'] <= 120)} landmarks")

Route Planning Insights:

• Quick Tours (≤30 min): 67 landmarks available
• Standard Tours (≤60 min): 89 landmarks available
• Extended Tours (≤120 min): 95 landmarks available
• Flexible Planning: Wide range of options for different time budgets

Data Quality Assessment

# Check for missing values and data consistency
print("Data Quality Assessment:")
print(f"Total landmarks: {len(landmarks_data)}")

# Check for missing fields
missing_fields = {
    'id': sum(1 for lm in landmarks_data if not lm.get('id')),
    'name': sum(1 for lm in landmarks_data if not lm.get('Landmark_name')),
    'class': sum(1 for lm in landmarks_data if not lm.get('Class')),
    'rating': sum(1 for lm in landmarks_data if lm.get('rating') is None),
    'geocoord': sum(1 for lm in landmarks_data if not lm.get('geocoord')),
    'dwell_time': sum(1 for lm in landmarks_data if lm.get('time taken to explore') is None),
    'io_type': sum(1 for lm in landmarks_data if not lm.get('indoor/outdoor'))
}

print("Missing values by field:")
for field, count in missing_fields.items():
    print(f"  {field}: {count} missing")

# Check data consistency
print(f"\nData Consistency:")
print(f"• All landmarks have unique IDs: {len(set(lm['id'] for lm in landmarks_data)) == len(landmarks_data)}")
print(f"• Rating range valid (0-5): {all(0 <= lm['rating'] <= 5 for lm in landmarks_data)}")
print(f"• Dwell time positive: {all(lm['time taken to explore'] > 0 for lm in landmarks_data)}")
print(f"• Valid indoor/outdoor: {all(lm['indoor/outdoor'] in ['indoor', 'outdoor'] for lm in landmarks_data)}")

Data Quality Results:

• Completeness: 100% - No missing values in any field
• Consistency: 100% - All data follows expected formats
• Uniqueness: 100% - All landmarks have unique identifiers
• Validity: 100% - All values within expected ranges

Recommendations for Dataset Users

1. For Route Optimization: Focus on 10-60 minute dwell times for realistic tours
2. For Recommendation Systems: Consider category preferences and rating distributions
3. For Mobile Apps: Balance indoor/outdoor landmarks based on weather conditions
4. For Educational Tools: Leverage high-rated cultural and research landmarks
5. For Time-Constrained Users: Use the dwell time analysis for efficient planning

Usage

To use this dataset, you can download the landmarks.json file and load it using a JSON parser in your preferred programming language.

import json

# Load the dataset
with open('landmarks.json', 'r') as f:
    landmarks_data = json.load(f)

print(f"Loaded {len(landmarks_data)} landmarks.")

# Example: Find all cultural landmarks
cultural_landmarks = [
    lm for lm in landmarks_data 
    if 'Culture' in lm.get('Class', [])
]

print(f"Found {len(cultural_landmarks)} cultural landmarks.")

# Example: Find high-rated indoor landmarks
high_rated_indoor = [
    lm for lm in landmarks_data 
    if lm.get('rating', 0) >= 4.5 and lm.get('indoor/outdoor') == 'indoor'
]

print(f"Found {len(high_rated_indoor)} high-rated indoor landmarks.")

Dataset Statistics

• Total Landmarks: 100+
• Categories: 7+ different classes (Culture, Research, Recreation, etc.)
• Rating Range: 0.0 - 5.0
• Dwell Time Range: 5 - 480 minutes
• Indoor/Outdoor Split: Balanced representation
• Geographic Coverage: Complete CMU campus

Related Work

This dataset is used in the following applications:

• CMU Explorer Space: https://huggingface.co/spaces/ysakhale/Tartan-Explore
• Content-Based Recommendation System: For personalized landmark recommendations
• Rating Prediction Model: For validating and predicting landmark ratings
• ML Route Optimization: For intelligent route planning

Citation

@misc{cmu-landmarks-dataset,
  title={CMU Landmarks Dataset},
  author={Yash Sakhale, Faiyaz Azam},
  year={2025},
  url={https://huggingface.co/datasets/ysakhale/Tartan-Explore}
}

Contact

For questions about this dataset, please refer to the CMU Explorer Space or the project repository.