Data Access and Usage

This document explains how to obtain and use the data required for the FreTIS project.

⚠️ Important Notice

Raw camera images are NOT included in this repository due to:

File size constraints (GitHub limits)
Copyright and usage restrictions (Caltrans data)
Privacy considerations (potential PII in images)

To use this project, you will need to obtain the data separately.

📸 Raw Camera Images

Source

Images are collected from Caltrans District 8 freeway cameras monitoring:

Interstate 10 (I-10)
Interstate 15 (I-15)
Interstate 215 (I-215)
State Route 60 (SR-60)

Camera Locations

Camera ID	Location Description	Approximate Address
`i1012westofhaven`	I-10 @ 2 West of Haven	I-10, San Bernardino County
`i101810mileswestofcherry`	I-10 @ 18/10 Miles West of Cherry	I-10, Riverside County
`i10200i210`	I-10 @ 200/I-210 Interchange	I-10/I-210, San Bernardino
`i1026westofrancho`	I-10 @ 26 West of Rancho	I-10, San Bernardino County
`i1033eastofwaterman`	I-10 @ 33 East of Waterman	I-10, San Bernardino County
`i1587foothillblvd`	I-15 @ 87 Foothill Blvd	I-15, Rancho Cucamonga
`i215207vinestreet`	I-215 @ 207 Vine Street	I-215, Riverside
`sr60153eastofeuclidavemuebonviewavenue`	SR-60 @ 153 East of Euclid	SR-60, Ontario
`sr60359westofvalleyway`	SR-60 @ 359 West of Valley Way	SR-60, Riverside County
`sr60362marketstreet`	SR-60 @ 362 Market Street	SR-60, Riverside

Total Images Used in Study: ~19,543 images (as of project date)

How to Obtain Images

Option 1: Scrape from Caltrans (Recommended for Replication)

Use the provided scraper script (if available in repository)
Access Caltrans QuickMap: https://quickmap.dot.ca.gov/
PeMS Camera Archive: http://pems.dot.ca.gov/ (requires account)

# Example scraper usage (adjust based on actual script)
python scripts/scrape_caltrans_cameras.py --cameras all --duration 30days

Option 2: Request from Research Team

For the exact dataset used in this study:

Contact: Ha Trinh (helentrinh9784@gmail.com)
Contact: Dr. Bilal Khan (bilal.khan@csusb.edu)
Note: Access may be restricted; approval required

Option 3: Use Your Own Images

The system can process any traffic camera images:

Place images in data/raw_images/[location_name]/
Follow naming convention: {camera_id}_{timestamp}.jpg
Ensure images are clear enough for vehicle segmentation

Data Structure

data/
└── raw_images/
    ├── i1012westofhaven/
    │   ├── image_20210101_0800.jpg
    │   ├── image_20210101_0900.jpg
    │   └── ...
    ├── i101810mileswestofcherry/
    │   └── ...
    └── [other camera locations]/
        └── ...

Usage Rights and Attribution

Caltrans Data Policy:

Images are from California Department of Transportation (Caltrans)
Check Caltrans terms of use: https://dot.ca.gov/
Attribution required: "Camera imagery courtesy of Caltrans"
For research use: Contact Caltrans District 8

Citation for This Dataset:

@dataset{FreTIS2025Data,
  author = {Khan, Bilal and Trinh, Ha},
  title = {Inland Empire Freight Traffic Camera Images},
  year = {2025},
  publisher = {California State University, San Bernardino},
  note = {Images courtesy of Caltrans District 8}
}

🔍 Processed Segmentation Data

What's Included

Sample outputs from OpenSeeD segmentation are included in:

OpenSeeD/extracted_data_full/samples/
sample_truck_segmentation.json
sample_truck_segmentation_20251201.json

Full Dataset

Due to size, full processed data is not included. To generate:

# Run segmentation pipeline (adjust based on actual implementation)
python openseed_segment.py --input data/raw_images/ --output OpenSeeD/extracted_data_full/

Alternatively, request full processed dataset from research team.

📊 Integrated External Datasets

The following datasets are integrated but not included in this repository:

1. CalEnviroScreen 4.0

Source: California OEHHA
URL: https://oehha.ca.gov/calenviroscreen/report/calenviroscreen-40
Data: Pollution burden, population characteristics, health indicators
License: Public domain (California state data)

2. EPA EJScreen

Source: U.S. Environmental Protection Agency
URL: https://www.epa.gov/ejscreen
Data: Environmental justice indicators
License: Public domain (U.S. federal data)

3. U.S. Census ACS (American Community Survey)

Source: U.S. Census Bureau
URL: https://www.census.gov/programs-surveys/acs
Data: Demographics, income, housing
License: Public domain

4. California Healthy Places Index

Source: Public Health Alliance of Southern California
URL: https://www.healthyplacesindex.org/
Data: Health indicators by census tract
License: Check website for terms

5. SWITRS/TIMS Crash Records

Source: California Highway Patrol (via UC Berkeley TIMS)
URL: https://tims.berkeley.edu/
Data: Traffic collision data
License: Requires registration and data use agreement

6. SCAG Warehouse Data

Source: Southern California Association of Governments
URL: https://scag.ca.gov/
Data: Warehouse locations, truck routes
License: Check SCAG data portal terms

7. LODES Employment Data

Source: U.S. Census Bureau LEHD
URL: https://lehd.ces.census.gov/data/
Data: Employment patterns
License: Public domain

🔐 Privacy and Ethics

Image Privacy Considerations

Traffic camera images may contain:

License plates (potentially identifiable)
Vehicles that could be tracked
Timestamps and locations

If using this system:

✅ Blur or redact license plates if publishing images
✅ Aggregate data to prevent individual tracking
✅ Follow institutional research ethics guidelines
✅ Do not use for surveillance or tracking purposes

Data Use Agreement

By using this data/code, you agree to:

✅ Use for research and educational purposes only
✅ Cite the original research (see README.md)
✅ Respect privacy and ethics guidelines
✅ Comply with all data provider terms of use
✅ Not use for commercial purposes without permission

📥 Quick Start (Without Raw Images)

You can explore the project without downloading all images:

Review sample outputs:
- sample_truck_segmentation*.png (example visualizations)
- sample_truck_segmentation*.json (example JSON output)
Understand methodology:
- Read LTC Proposal Freight Services_BK.docx.md
- Review analysis notebooks (if available)
Test with your own images:
- Place a few test images in data/raw_images/test/
- Run segmentation pipeline
- Verify results

🤝 Data Sharing Requests

For Researchers

If you need access to the full dataset for research:

Contact:

Ha Trinh: helentrinh9784@gmail.com
Dr. Bilal Khan: bilal.khan@csusb.edu
Subject line: "FreTIS Data Access Request"

Include in your request:

Your research purpose
Institutional affiliation
IRB approval (if applicable)
Planned use and analysis

For Collaborators

Active collaborators may access shared data storage:

CSUSB institutional storage
Secure shared drive (access by permission)

❓ Frequently Asked Questions

Q: Can I use different camera sources?
A: Yes! The pipeline works with any traffic camera images. Just organize them in the same directory structure.

Q: How much disk space do I need?
A: For ~19,543 images (~300KB each): ~5-6 GB for raw images, plus ~2-3 GB for processed outputs.

Q: Do I need GPU for segmentation?
A: Recommended but not required. GPU significantly speeds up OpenSeeD inference.

Q: Can I use a subset of cameras?
A: Yes! Start with one or two camera locations to test the pipeline.

Q: Are the images time-stamped?
A: Yes, timestamps are part of the filename/metadata. See scraper script for details.

📞 Questions?

Data access: Contact research team (emails above)
Technical issues: Open an issue on GitHub
Caltrans permissions: Contact Caltrans District 8 directly
General inquiries: See README.md

Last Updated: January 2025
Maintained by: Ha Trinh, Dr. Bilal Khan
Institution: California State University, San Bernardino

FilesExpand file tree

DATA.md

Latest commit

History