rlogh/negativesirl

CMU Campus Negative Images Dataset

Dataset Description

This dataset contains negative example images collected around Carnegie Mellon University campus in Pittsburgh, Pennsylvania. These images represent non-lanternfly subjects that were used as negative training data for binary classification models. The dataset was created as part of an academic project (Project 1) focused on developing machine learning models for invasive species identification.

Dataset Summary

The CMU Campus Negative Images Dataset consists of carefully curated images that do not contain spotted lanternflies, collected from various locations around the CMU campus. These images serve as negative examples to train binary classification models to distinguish between lanternflies and other subjects.

The dataset includes:

• Diverse Negative Examples: Images of various insects, objects, and environmental features
• Campus Context: Images collected from CMU campus environments
• Training Purpose: Specifically curated for machine learning model training
• Quality Control: Images verified to not contain spotted lanternflies

This dataset supports the development of robust binary classification models by providing comprehensive negative examples that help prevent false positive identifications.

Supported Tasks and Leaderboards

• Image Classification: Binary classification (Lantern Fly vs Non-Lantern Fly)
• Object Detection: Identification of non-lanternfly subjects
• Model Training: Negative example training data for classification models
• Computer Vision: General object recognition and classification

Dataset Structure

Data Instances

Each data instance contains:

• image: High-resolution photograph (typically 512x512 or higher)
• filename: Unique identifier for the image file
• category: General category of the negative example (e.g., "other_insect", "environment", "object")

Data Fields

Field	Type	Description
image	Image	High-resolution photograph of non-lanternfly subject
filename	String	Unique filename identifier
category	String	General category of the negative example

Data Splits

Split	Examples	Description
train	~80%	Training data for model development
validation	~10%	Validation data for model evaluation
test	~10%	Test data for final model assessment

Dataset Creation

Curation Rationale

This dataset was created to:

1. Support Binary Classification: Provide negative examples for training lanternfly vs non-lanternfly classifiers
2. Improve Model Robustness: Help models learn to distinguish lanternflies from similar-looking subjects
3. Prevent False Positives: Reduce incorrect lanternfly identifications
4. Academic Research: Support machine learning model development in academic context

Source Data

• Collection Method: Systematic photography around CMU campus
• Geographic Scope: Carnegie Mellon University campus and surrounding areas
• Subject Types: Various insects, environmental features, and objects
• Quality Standard: Images verified to not contain spotted lanternflies

Data Collection Process

1. Field Collection: Systematic photography of non-lanternfly subjects
2. Category Classification: Grouping images by general subject type
3. Quality Verification: Confirming absence of spotted lanternflies
4. Data Organization: Structuring for machine learning training
5. Validation: Cross-checking image quality and categorization

Image Categories

The dataset includes various categories of negative examples:

• Other Insects: Various insect species found on campus
• Environmental Features: Leaves, bark, ground cover, etc.
• Objects: Campus infrastructure, debris, etc.
• Background Elements: Various environmental backgrounds

Personal and Sensitive Information

No personal or sensitive information is included in this dataset. All images contain only environmental subjects and objects from public campus areas.

Considerations for Using the Data

Social Impact of Dataset

This dataset supports:

• Machine Learning Research: Development of robust classification models
• Academic Education: Teaching materials for computer vision and ML
• Environmental Monitoring: Improving automated species identification
• Model Validation: Testing and validating classification algorithms

Discussion of Biases

Potential biases in the dataset include:

• Geographic Bias: Limited to CMU campus environment
• Seasonal Bias: Collection timing may favor certain seasons
• Environmental Bias: Campus-specific environmental conditions
• Collection Bias: Photographer preferences and collection methods
• Temporal Bias: Limited to specific time periods

Other Known Limitations

• Limited geographic scope (single campus)
• Potential underrepresentation of certain subject types
• Variable lighting and environmental conditions
• No geolocation data (images only)

Additional Information

Dataset Curators

• Primary Curators: Student Researchers
• Institution: Carnegie Mellon University
• Course: Project 1
• Academic Year: 2024

Licensing Information

This dataset is released under the Creative Commons Attribution 4.0 International License (CC BY 4.0).

Citation Information

@dataset{cmu_negative_images_2024,
  title={CMU Campus Negative Images Dataset: Negative training examples for spotted lanternfly classification},
  author={Student Researchers},
  year={2024},
  institution={Carnegie Mellon University},
  course={Project 1},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/rlogh/negativeirl},
  license={CC BY 4.0}
}

Academic Context

This dataset was created as part of an academic project focused on:

• Machine Learning Model Development: Training binary classification models
• Computer Vision Applications: Image classification and object detection
• Environmental Data Collection: Systematic field data collection
• Model Training: Providing comprehensive negative examples

Usage Examples

from datasets import load_dataset
import matplotlib.pyplot as plt

# Load the dataset
dataset = load_dataset("username/cmu-negative-images")

# Display sample images
fig, axes = plt.subplots(2, 3, figsize=(12, 8))
for i, (image, category) in enumerate(zip(dataset['train']['image'][:6], 
                                        dataset['train']['category'][:6])):
    row, col = i // 3, i % 3
    axes[row, col].imshow(image)
    axes[row, col].set_title(f"Category: {category}")
    axes[row, col].axis('off')

plt.tight_layout()
plt.show()

Training Integration

This dataset is designed to be used alongside positive lanternfly examples for training binary classification models

Quality Assurance

All images in this dataset have been:

1. Manually Verified: Confirmed to not contain spotted lanternflies
2. Quality Checked: Assessed for image quality and clarity
3. Categorized: Grouped by general subject type
4. Validated: Cross-checked for accuracy

Acknowledgments

• Carnegie Mellon University for providing the academic framework
• Course instructor for project guidance
• CMU campus community for access to collection sites
• The machine learning community for classification methodologies

Future Work

Potential extensions of this dataset include:

• Expanded Categories: More diverse negative example categories
• Geographic Expansion: Collection from additional locations
• Temporal Extension: Multi-season data collection
• Metadata Enhancement: Additional categorization and annotation

---

This dataset represents academic work in machine learning model development and environmental data collection at Carnegie Mellon University.