Global Computer Vision Leaderboard

Scylla Technologies: Independently Benchmarked Computer Vision

In the rapidly evolving world of artificial intelligence and computer vision, benchmark results often tell only part of the story. We track independent benchmark results because prevention technology should be judged on evidence, not marketing claims.

The Global Standings

COCO test-dev benchmark is one of the most recognized evaluation platforms for computer vision. Scylla publishes its ScyllaNet results on the public COCO leaderboard at CodaLab, where the standings can be read directly. We point to the source rather than recite a placement. Standings change, and the published record is the authority. See scylla.ai for current results. Benchmark placement is only part of a procurement decision. Federal law also governs whose cameras a buyer may install. Section 889 of the Fiscal Year 2019 National Defense Authorization Act prohibits federal agencies, contractors, and federal grant recipients from procuring covered telecommunications and video-surveillance equipment produced by Huawei, ZTE, Hytera, Hikvision, or Dahua, including their subsidiaries and affiliates, where that equipment is used for public safety, security of government facilities, physical security surveillance of critical infrastructure, or other national security purposes. For federal grant recipients the restriction took effect on August 13, 2020, under 2 CFR §200.216.

That matters to any school, agency, or contractor spending federal dollars on security technology. Compliance is not optional and it is checked at award. That is where Scylla fits, as a computer-vision solution available to U.S. and NATO-aligned enterprises.

Why This Matters

AI is no longer just about accuracy. It’s about trust, alignment, and operational readiness. Companies and governments across the U.S. and Europe need solutions that are not only state-of-the-art but also deployable within trusted regulatory and geopolitical frameworks.

• Compliance & Trust: Scylla’s models can be deployed without the restrictions that come with foreign adversarial technologies.

• Efficiency at Scale: With only ~40 million parameters, ScyllaNet matches the accuracy of models requiring 300 million to 3 billion parameters, making it dramatically faster and cheaper to deploy.

• Security First: Scylla is built with national security, enterprise readiness, and public safety use cases in mind.

How ScyllaNet Is Measured

ScyllaNet pairs competitive accuracy with high throughput. It achieves:

• 0.66 mAP, on par with the global leaders.

• Top recall scores, especially in small-object detection.

• 90×–200× faster performance than heavyweight architectures like Co-DETR ViT-L and InternImage-G.

For industries where milliseconds matter, such as defense, public safety, critical infrastructure, and healthcare, Scylla’s balance of speed and accuracy fits those demands.

The Power of Computer Vision in Prevention

Scylla’s impact goes far beyond benchmarks. Its computer vision platform powers real-world prevention:

• Gun Detection: Real-time identification of visible firearms in schools, businesses, and public spaces, giving first responders a critical early warning to stop violence before it escalates.

• Autonomous Surveillance Platforms: Scylla integrates with drones, fixed cameras, and IoT networks to provide always-on monitoring — delivering continuous situational awareness without human fatigue.

• Object Recognition Through Neural Networks: From suspicious items in airports to unattended bags in stadiums, Scylla’s deep neural networks classify threats accurately, even in complex, crowded environments.

These capabilities make Scylla not just a technology provider, but a force multiplier for prevention and public safety.

Proud Partner of the Active Shooter Prevention Project

Scylla Technologies is honored to be a proud partner of the Active Shooter Prevention Project (ASPP™), aligning advanced AI innovation with a national mission: making communities as prepared against active shooter threats as they are against fire emergencies.

Together, Scylla and ASPP™ are setting a new prevention-first standard, where technology, training, and trusted partnerships converge to protect lives and restore freedom from fear.

The Road Ahead

The implications are clear: Scylla Technologies is a computer vision provider available to allied nations and U.S. businesses today, and its results are published on the public COCO leaderboard at CodaLab, where the standings can be read directly.

As AI regulation, supply chain security, and global competition tighten, Scylla’s rise signals a new standard: world-class performance without compromise.

🚀 Scylla Net  Faster. Smarter. Aligned with the future.

---

Where this fits in prevention: Detection technology is one input to a prevention programme, never a substitute for one. Software can surface a signal; only a trained organization with a defined route can act on it in time.

Understanding how these events unfold is what makes prevention possible. → The P.R.O. Model™

See how we deliver P.R.O. Model™ workshops → · Request a consultation