AI system
- System
-
Rapiscan 3D X-ray CT marine-wildlife detection algorithms (3D convolutional neural networks with 3D Threat Image Projection)
- Model
-
Experimental three-dimensional convolutional neural networks.
- Release
-
The experimental convolutional networks have no published release identifier. [1]
- Application developer
-
Rapiscan Systems Ltd with Taronga Conservation Society Australia and Macquarie University
- Role
-
The networks identified marine-wildlife specimens in CT images of baggage.
- Configuration
-
Researchers trained the networks on CT scans and synthetic images with inserted marine-wildlife specimens.
- Inputs and outputs
-
The networks identified marine-wildlife specimens in CT images of baggage.
- Autonomy
-
Not assessed
- Workflow
-
Researchers scanned marine specimens with a Rapiscan RTT110. They inserted specimens digitally into other CT images to expand the training data and then tested the detection algorithms. [1]
- Human oversight
-
Not assessed
- Relation to AI
-
AI classified objects in CT baggage scans. Most positive test bags used synthetically inserted specimens, and testing drew from the same specimen population. The paper is original research with disclosed Rapiscan involvement, not an independent deployment evaluation. [1]
- AI contribution
-
AI classified objects in CT baggage scans. Most positive test bags used synthetically inserted specimens, and testing drew from the same specimen population. The paper is original research with disclosed Rapiscan involvement, not an independent deployment evaluation. [1]
Benefits
- Outcome identifiers
-
Not assessed
- Benefit
-
The algorithms detected 95% of shark fins, 96% of seahorses, and 86% of sea cucumbers in the tests. The overall false-alarm rate was 13%. [1]
- Category
-
Improved Task Accuracy
- Status
-
Not assessed
- Beneficiary groups
-
People. Organizations. Society. [1]
- Beneficiaries
-
Not assessed
- Reach
-
Not assessed
- Distribution and access
-
Not assessed
- Duration
-
Not assessed
- Expected?
-
Yes [1]
- Costs and harms
-
Not assessed
Measurement
- Studies
-
Not assessed
- Study design
-
Not assessed
- Measurement
-
Not assessed
- Setting and sample
-
Not assessed
- Measures
-
detection rates of 95% (shark fin), 96% (seahorse) and 86% (sea cucumber)
- Comparison
-
known scanned samples (ground truth), not human inspectors
- Results
-
Result: PD shark fin: 95%, seahorse: 96% and sea cucumber: 86%.
- Additional value from AI
-
Not assessed
- Limitations
-
Not assessed
- Denominators
-
Not assessed
- Materials for replication
-
Not assessed
Money
- Kind of value
-
Not assessed
- Amount
-
Not assessed
- Currency and prices
-
Not assessed
- Period and unit
-
Not assessed
- Realization
-
Not assessed
- Recipients and payers
-
Not assessed
- Gross or net
-
Not assessed
- Included costs
-
Not assessed
- Comparison
-
Not assessed
- Method
-
Not assessed
- Return and payback
-
Not assessed
People and places
- Participants
-
Developer: Rapiscan Systems Ltd with Taronga Conservation Society Australia and Macquarie University. Operator: None; proof of concept with no customs authority deployment.
- Jurisdictions
-
Benefit: Sydney, New South Wales,
Australia (research setting). Developer: 
United Kingdom. - Ownership and control
-
Not assessed
- Dependencies
-
Not assessed
- Cross-border connections
-
Not assessed
- Geopolitical implications
-
Not assessed
Follow-up
- Conditions for replication
-
Customs services would need suitable CT scanners, reliable performance on concealed and mixed consignments, and a lawful process for investigating alerts. [1]
- Replication attempts
-
Not assessed
- Later findings
-
The authors say implementation at the front lines is the next step; no customs trial or deployment of the algorithms is reported in the cited sources.
- Availability
-
Not assessed
- Institutional decision
-
Not assessed
- Grounds
-
Not assessed
- Action and result
-
Not assessed
- Success criterion
-
Not assessed
Evidence
- Source numbers
-
AIBD-0362-source-0001 [1]
- Citations
-
Numbered references below. [1]
- Language
-
Not assessed
- Source relationships
-
Developer [1]
- Original evidence
-
The final paper supplies the detection rates in table 3. Its supplement was inaccessible, and raw data require permission. The training total in § 3.1 appears inconsistent with the listed class counts. [1]
- Access and versions
-
The full paper is available, including its methods, tables, and data statement. Data_Sheet_1.docx is named as a supplement but was inaccessible. [1]
- Supported findings
-
Not assessed
- Evidence links
-
Not assessed
- Evidence strength
-
Not assessed
- Missing evidence
-
Not assessed
Grades
- Assessed benefits
-
People. Organizations. Society. [1]
- Depth
-
Not assessed
- Reasons
-
The study measures detection in CT scans, mainly with specimens inserted digitally into test bags. It reports no customs seizure, operational evaluation, or individual benefit. [1]
- Observed or intended
-
Not assessed
- Rubric
-
Not assessed
- Assessor and date
-
Not assessed
- Alternative assessments
-
Not assessed
- Fill
-
The makers' claim
Entry details
- Kind
-
Not assessed
- Related entries
-
Not assessed
- Field
-
Wildlife trafficking detection at borders. Security-screening vendor and conservation research partnership.
- Purpose
-
The researchers tested whether AI could detect marine wildlife in CT scans of mail and luggage to assist customs officers. [1]
- Dates
-
Source published: 8 June 2026. Updated: 24 October 2026. [1]
- Stage
-
Not assessed
- Version
-
SHA-256: ad828aba7302…
References
- Other identifiers
-
Not assessed
- Category mappings
-
Not assessed
- Brief
-
Not assessed
- Revisions
-
7 revisions. Latest: 24 October 2026. [1]
Sources
-
Peer-reviewed study Developer
Pirotta, V., Meagher, P., Phan, H. T. H., Shen, K., Thompson, W., Dolman, B., Bending, Z., Smoothey, A. F., Hay, A., Rowe, C. E., Motamayor, A., Liddell, B., Allende, A., Allman, B. E., O'Brien, J. K. (2026). Marine wildlife trafficking: use of AI algorithms for the autodetection of shark fins, seahorses and sea cucumbers. Frontiers in Ocean Sustainability. doi:10.3389/focsu.2026.1776978; §§2–3, tables 1–3, data availability, and conflict-of-interest statement.
The authors produced the scanning experiment. Rapiscan employees participated and the company funded equipment, analysis, interpretation, and writing. Peer review does not make this an independent replication.