Currently ALERT has three active datasets available to the research community that have developed out of transition tasks supported by DHS. The datasets which are currently available are related to ALERT’s CT Segmentation Initiative, Reconstruction Initiative, Automated Threat Recognition (ATR) Initiative and ALERT Video Analytics Re-Identification work. If you would like to inquire about accessing this data, please complete our Dataset Request Form.
The purpose of the data we have procured is to provide security-like data to academic researchers, to discover and evaluate the present state-of-the-art, and to stimulate additional communication and research in the algorithm research community.*
* Our data (except for the Airport Re-Identification Dataset) is available to researchers under a Non-Disclosure Agreement(NDA). The NDA states that the data is owned by ALERT. If you choose to receive this data, you will not receive or create SSI data. Any publications generated, including reports and presentations, must be reviewed and approved by the Northeastern University/ALERT Research Evaluation Advisory Panel (REAP) in advance of publication or dissemination.
ALERT has leveraged the advances of medical CT, and contracted with a vendor to obtain representative datasets of packed luggage and reference objects. Contractual arrangements were made to scan luggage on a state-of-the-art medical CT scanner at the manufacturer’s factory. Approximately 900 objects were fully segmented from 62 luggage datasets to span the spectrum of packing, density, arrangement, orientation, and size difficulty.
ALERT personnel performed the following steps:
- Procured luggage and items to pack in them.
- Labeled, photographed and cataloged the items.
- Packed the suitcases
- Created a database of items as packed into suitcases
- Created video tapes of unpacking the luggage
- Scanned the luggage at the vendor’s factory
The vendor performed the following steps:
- Reconstructed the projection data corresponding to the scans of the luggage. The reconstructions were performed with offline reconstruction. The resulting resolution was approximately 3 mm FWHM.
- Converted the images to DICOM format.
ALERT then performed the following additional steps.
- Converted the DICOM images to TIFF files.
- Used a network in Mevislab to semi-automatically outline the items in the scans. The resulting data was known as ground truth data, label images and AO images.
- Divided the scans into four sets denoted: qualification, training, validation and evaluation.
- Distributed the data per instructions from the leadership team.
- Revised the data based on feedback from various stakeholders.
Reconstruction Initiative Dataset
ALERT created a database of projection and image data corresponding to scans of objects of interest in the presence of various amounts of clutter, using a medical CT scanner. Reconstructed images using filtered back-projection that match the images obtained on the medical scanner were used to generate the database.
A database was created of the following items:
- Projection data (raw and corrected) and images.
- Meta‐data including scanner geometry and file formats.
- Maps of objects in the CT scans known as ground truth data.
- Descriptions of the items themselves and how they were packed.
Simulated packed bags containing mathematical phantoms of objects commonly found in suitcases were created. The simulated projections closely match the medical scanner.
The following objects of interest were chosen for the researchers to evaluate.
- Saline with varied concentrations
- Rubber sheets
- Glass + plastic beads
These materials were chosen to be a set of reproducible objects. Other clutter objects were also scanned as well as calibration objects. The beads were later dropped from consideration because the CT values of the glass were clipped in the reconstructed images.
Automated Threat Recognition(ATR) Initiative
ALERT Airport Re-Identification Dataset
As part of the ALERT video analytics effort, researchers at Northeastern University and Rensselaer Polytechnic Institute developed an annotated dataset that accurately reflects the real-world person re-identification problem. The dataset was constructed using video data from the six cameras installed post central security checkpoint at an active commercial airport within the United States. (No NDA required)
Innovation Fast Track
EXCLUSIVE TO CENTER MEMBERS: Log in to explore your Innovation Fast Track, an online portal that gives you one-stop access to center researchers, funding opportunities, exclusive events and more.Learn More
Students, get connected with Industry Partners, find out about research and volunteer opportunities within ALERT, or get advice on how to best showcase your work at Student Central.Learn More
ALERT is led by Northeastern University (NEU) and includes three key academic strategic partners: Boston University, Purdue University, and University of Rhode Island (URI).Learn More