After the initial submittals, respondents will present the capabilities of their specific software to the MSU Department of Police and Public Safety Staff. Specific expectation for the presentation include: 1. Accuracy a. Accurately identifying a weapon when it is present. Leading systems often achieve high accuracy rates, such as 95–99%, for clearly visible weapons. b. Minimizing false alarms when an object is not a weapon. c. Availability of visible confidence level shown for each alarm d. Reliable long-range operation, detecting weapons from a distance, dependent on camera resolution. 2. Speed a. Low Latency- processing video almost instantaneously b. High frame rate- to avoid missing threats 3. Robustness- perform reliably under challenging conditions a. Environmental resilience- various lighting conditions low-light, shadows and weather b. Visual Clutter- campus atmosphere heavy traffic at times 4. Weapon recognition capability- The system should be able to identify a wide range of weapons, including handguns, rifles, shotguns, and knives. 5. Integration with existing systems a. Camera compatibility: Working with standard Internet Protocol (IP) cameras of multiple manufacturers and Genetec (VMS). b. Multi-camera support: Monitoring and analyzing feeds from an entire network of cameras. 6. Deployment and infrastructure: Implementation must consider the unique needs of the environment. a. Coverage planning: Ensuring security cameras cover all high-traffic and high-risk areas, like entry points, to eliminate blind spots. b. Edge and cloud processing: The system should be capable of both on-device (edge) and cloud-based analysis for rapid processing and complex scenario handling where possible. 7. Threat assessment: Beyond simple detection, advanced systems analyze behavior and context to reduce false positives. This may include: a. Using contextual analysis to distinguish between harmless items and actual threats. b. Requiring human verification of alerts for complex or ambiguous scenarios
