A human being watching a bank of security monitors loses meaningful attention after about twenty minutes, a finding security researchers have documented for decades and that anyone who’s actually pulled a monitoring shift can confirm from direct experience. Video anomaly detection software exists specifically to solve that attention problem: instead of a person scanning dozens of feeds hoping to catch the one moment something unusual happens, the system watches continuously and flags the specific clip, a person crossing a perimeter line, an object left unattended, a crowd forming somewhere it shouldn’t, for a human to actually review. The category has matured considerably, and it’s also seen real consolidation and at least one high-profile withdrawal worth knowing about before evaluating vendors.

A correction worth making before anything else

IBM Video Analytics, a name that appeared in nearly every version of this comparison for years, is no longer sold or supported. IBM formally discontinued support as of April 30, 2023, and its own product withdrawal documentation states plainly that there is no direct replacement offered. This followed IBM’s earlier 2019 withdrawal of its Intelligent Video Analytics product specifically over facial recognition and human rights concerns raised by media scrutiny, a decision IBM made publicly and deliberately rather than a quiet product sunset. Anyone evaluating this category today should treat any reference to “IBM Video Analytics” as historical rather than current, and look toward the vendors below, several of whom have absorbed the market share IBM’s exit left open.

BriefCam, now part of Canon

BriefCam was acquired by Canon in 2018 and now operates as part of Canon’s broader video surveillance portfolio, and that backing has translated into real, continued investment in the platform rather than the stagnation that sometimes follows an acquisition. Its core strength remains video content analytics that make existing footage genuinely searchable rather than something that only gets reviewed after an incident: filtering hours of recorded video by object type, color, direction of movement, or specific behavior compresses what used to be a manual review of an entire day’s footage into minutes. Its real-time anomaly detection layer sits on top of that same search infrastructure, flagging unusual patterns as they happen rather than only after the fact.

Recent platform updates have specifically emphasized generative AI readiness for deeper natural-language search and analysis, extending the “search your footage like a database” concept the platform built its reputation on. For enterprise security operations that need genuine forensic search alongside live anomaly alerting, it remains one of the strongest, best-resourced options in the category following its Canon acquisition.

Avigilon, part of Motorola Solutions

Avigilon, now under Motorola Solutions following that 2018 acquisition, combines video analytics with the broader Motorola security ecosystem, license plate recognition, access control, and unified alarm management, rather than operating as a standalone analytics layer. Its Focus of Attention feature specifically highlights unusual events for a security operator’s review rather than requiring someone to manually scan every camera feed, and its self-learning video analytics improve detection accuracy over time as the system observes more of a specific site’s normal patterns.

Being part of Motorola’s larger Alta and Unity platform ecosystem means the deepest value shows up for organizations adopting Motorola’s broader security stack rather than Avigilon’s analytics as an isolated point solution. For enterprises already invested in or considering that ecosystem, the integration depth is a genuine advantage over a standalone analytics vendor.

Milestone Systems, XProtect

Milestone’s XProtect platform approaches video anomaly detection from a genuinely open, vendor-agnostic angle: rather than building a single proprietary analytics engine, XProtect functions as an open platform that third-party analytics providers, including BriefCam’s engine among others, plug directly into through Milestone’s App Platform. That openness matters for organizations that don’t want to be locked into a single vendor’s analytics roadmap and would rather choose best-of-breed analytics modules while keeping video management infrastructure consistent underneath.

That flexibility asks more of an integrator or in-house team to actually assemble a complete analytics solution from separate components rather than buying one fully bundled product. For organizations with the technical capacity to make that assembly work, or working with an integrator who specializes in Milestone deployments, the resulting system can be more precisely tailored than a single-vendor bundle offers.

Agent Vi, part of Irisity

Agent Vi, acquired by Swedish AI security company Irisity in a 2021 deal valued around $67.5 million, brought deep learning-based anomaly detection into a larger, better-resourced organization following the acquisition. Its analytics cover behavioral pattern recognition, unauthorized access detection, and traffic violation monitoring, with strong scalability across multi-camera deployments through both on-premise and cloud architectures. Integration with a wide range of third-party video management systems means it can layer onto existing camera infrastructure rather than requiring a full hardware replacement.

As part of the broader Irisity organization, its product roadmap and branding are subject to the parent company’s strategic direction rather than operating fully independently, which is worth knowing if long-term product continuity under a specific name matters to a purchasing decision. For retail, traffic management, and general security applications needing solid behavioral analytics without committing to the biggest, most expensive enterprise platforms, it remains a genuinely capable mid-market option.

iOmniscient

iOmniscient targets genuinely complex, large-scale environments, airports, smart cities, major industrial facilities, where standard anomaly detection tuned for a single camera or small site breaks down under real operational scale. Its analytics cover abandoned object detection, crowd behavior analysis, and traffic pattern monitoring, with heavy customization built specifically for a given deployment’s unique environment rather than a one-size-fits-all default configuration. That customization depth is exactly what large, complex sites with unusual layouts and specific operational needs require.

The tradeoff for that customization is cost and implementation complexity well beyond what a smaller organization would need or could justify; this is genuinely enterprise-scale software built for enterprise-scale problems. For the specific niche of major public infrastructure and large industrial sites, it remains one of the more specialized, capable options.

Senstar Symphony

Senstar built its broader reputation on perimeter security and intrusion detection specifically, and Symphony, its video management and analytics platform, reflects that heritage by integrating video analytics tightly with perimeter sensors, fence detection, buried cable systems, radar, rather than treating video as an isolated data source. For critical infrastructure, industrial sites, and facilities where the actual security priority is detecting an intrusion at the perimeter before it becomes an interior problem, that integrated sensor-plus-video approach genuinely outperforms a video-only analytics platform.

Its strength in perimeter-specific use cases doesn’t necessarily extend to the broader behavioral and crowd analytics that a retail or public-space deployment might prioritize instead. For critical infrastructure and industrial security specifically, it remains a strong, purpose-built choice.

Viseum IMC

Viseum’s Intelligent Moving Camera pairs a distinctive 360-degree, multi-tasking camera hardware design with its own intelligent video analytics software, autonomously detecting and tracking suspicious activity across a wide field of view rather than relying on a fixed camera angle to catch everything. That autonomous tracking capability, following a detected anomaly across the camera’s full range of motion rather than requiring an operator to manually pan and follow, is a genuine differentiator for perimeter and public-space security specifically.

It’s a smaller, more specialized vendor than the major platforms above, with a narrower installed base and correspondingly less third-party integration and community support. For sites specifically suited to its distinctive camera-plus-analytics approach, airports, large public spaces, critical perimeters, it remains a genuinely distinctive option in a market otherwise dominated by fixed-camera analytics layered on top of conventional hardware.

Hikvision VCA

Hikvision’s Video Content Analysis suite ships built directly into Hikvision’s own camera and video management hardware, which means analytics capability comes bundled with the hardware purchase rather than requiring a separate software licensing relationship on top of existing cameras. Behavior analysis, object counting, and intrusion detection cover common retail, transportation, and general security use cases at a price point generally more accessible than the enterprise-focused platforms above, reflecting Hikvision’s broader positioning as a high-volume hardware manufacturer rather than a specialized analytics-only vendor.

Its analytics depth trails the more specialized platforms in this list for genuinely complex or unusual detection scenarios, and organizations in some regions and industries have faced procurement restrictions on Hikvision hardware specifically tied to broader geopolitical and supply-chain security policy, worth checking against your own organization’s procurement rules before committing. For straightforward security and retail analytics needs where hardware-bundled software keeps overall cost down, it remains a widely deployed, practical option.

IntelliVision, part of Nortek Security & Control

IntelliVision, acquired by Nortek Security & Control in 2018, continues operating as an active subsidiary with its deep learning-based video and audio analytics deployed across a genuinely large installed base, reportedly running on more than five million cameras worldwide. Its analytics cover loitering detection, suspicious activity recognition, and traffic monitoring, with broad compatibility across a wide range of third-party camera systems and video management platforms rather than requiring proprietary hardware.

As a Nortek subsidiary, its product direction sits within a larger corporate structure rather than operating as a fully independent company, similar to how Agent Vi now operates under Irisity. For organizations wanting broad camera compatibility and a genuinely mature, widely deployed analytics engine rather than a newer, less-proven platform, it remains a solid, established choice.

Edge processing versus cloud analysis, and why it matters for your bandwidth budget

A genuinely important technical decision underlying every platform above, and one that rarely gets its own line item in a sales conversation, is whether anomaly detection actually runs on the camera or a local server at the site, edge processing, or whether raw video streams to a cloud service for analysis there instead. Edge processing keeps bandwidth requirements dramatically lower, since only detected events and short clips need to travel off-site rather than continuous full-resolution video from every camera, which matters enormously for a large facility with dozens or hundreds of cameras on a limited network connection. Cloud-based analysis can offer easier centralized management and faster software updates across a distributed multi-site deployment, but it demands real, sustained bandwidth and raises additional data residency and privacy questions depending on where that cloud processing actually happens and which jurisdiction’s laws apply to data crossing a border in transit. Milestone’s open platform and most of the hardware-bundled options, Hikvision and Viseum among them, lean toward edge or on-premise processing by design, while some of the larger enterprise platforms offer a genuine choice between deployment models depending on a site’s specific bandwidth and data governance requirements. Asking a vendor directly where processing actually happens, not just where the management dashboard lives, is worth doing before signing a contract rather than discovering a bandwidth problem after a full camera rollout.

Choosing based on what your site actually needs

Large, complex sites, airports, smart city deployments, major industrial facilities, generally need iOmniscient’s or Senstar’s specialized depth over a general-purpose platform stretched beyond what it was built for. Organizations already committed to a specific hardware or platform ecosystem, Motorola’s broader security stack or Canon’s imaging portfolio, get real, compounding value from staying inside that ecosystem with Avigilon or BriefCam respectively rather than introducing a separate, disconnected analytics vendor. Sites prioritizing genuine vendor flexibility and best-of-breed component selection over a single bundled product should look toward Milestone’s open XProtect platform. And organizations with straightforward retail, general security, or traffic monitoring needs on a real budget constraint are generally well served by Hikvision’s hardware-bundled analytics or IntelliVision’s broadly compatible software layer, without needing the enterprise-scale depth, and enterprise-scale cost, of the more specialized platforms above.

Privacy and regulatory considerations that shape real deployment decisions

IBM’s 2019 withdrawal of its facial recognition-capable analytics product over human rights concerns wasn’t an isolated event; it reflects a genuinely active, ongoing regulatory and public scrutiny landscape around video analytics broadly, and particularly around facial recognition and biometric identification specifically. Several jurisdictions now restrict or require specific disclosure and consent frameworks for facial recognition and biometric analytics in ways that don’t apply to simpler anomaly detection, object counting, or behavior pattern recognition that doesn’t attempt to identify specific individuals. Before deploying any platform in this category, checking what analytics capabilities are actually enabled, and whether facial recognition or biometric identification features are part of the deployment or explicitly disabled, against your specific jurisdiction’s current legal requirements is a genuinely necessary step that a features comparison alone won’t cover. This is worth treating as a procurement requirement rather than an afterthought, since a platform capable of facial recognition doesn’t necessarily mean deploying that specific capability is legally permitted in every location.

Piloting before a full rollout, and why it’s worth the delay

Given how much anomaly detection accuracy depends on a system learning a specific site’s normal patterns, deploying a full platform across every camera on day one, before that learning process has had real time to mature, tends to produce exactly the false-positive fatigue problem described above at the worst possible moment, right when staff are forming their first impressions of whether the system is trustworthy. A more reliable rollout pattern runs a genuine pilot on a subset of cameras covering the highest-priority areas first, giving the system real weeks rather than days to adapt to that specific location’s normal traffic, lighting changes, and activity patterns before expanding coverage further. This also gives a security team the chance to catch integration issues, a camera angle that consistently confuses the analytics, a lighting condition that triggers false alerts at a predictable time of day, on a smaller, more manageable scale before those same issues repeat themselves across a much larger deployment. Vendors who resist a genuine pilot period in favor of pushing straight to a full contract are worth treating with some skepticism, since a platform confident in its own real-world performance should have no problem proving that performance on a limited scale first.

Frequently asked questions

Is IBM Video Analytics still available to purchase anywhere?

No. IBM discontinued support for IBM Video Analytics as of April 30, 2023, and its own withdrawal documentation states there is no direct replacement product. Organizations still running it should plan a migration to an actively supported platform, since continuing on unsupported software carries real security and reliability risk over time.

Do all of these platforms include facial recognition?

No, and it varies significantly by vendor and specific deployment configuration. Many of the platforms above are built primarily around behavior and object anomaly detection, unattended objects, unusual movement patterns, perimeter intrusion, rather than individual identification. Facial recognition, where it exists as a capability, is often a separately licensable or explicitly configurable module rather than something enabled by default, precisely because of the regulatory sensitivity around it.

Can these systems work with cameras I already have installed?

Most software-focused platforms, IntelliVision, Agent Vi, and Milestone’s open platform among them, are built for broad compatibility with existing third-party camera hardware. Hardware-bundled options like Hikvision’s VCA and Viseum’s IMC tie analytics more tightly to their own camera hardware, which means adopting them may involve at least partial hardware replacement rather than a pure software overlay onto an existing camera investment.

How much does false-positive fatigue affect these systems in practice?

It’s a genuine, ongoing operational concern across the entire category, not a solved problem for any single vendor. A system that generates too many false alerts trains security staff to start ignoring notifications entirely, which defeats the purpose of automated detection just as thoroughly as a system that misses real events. Most platforms have improved meaningfully on this front through machine learning that adapts to a specific site’s normal patterns over time, but budgeting a real tuning period after initial deployment, rather than expecting accurate detection immediately out of the box, is a realistic expectation regardless of which platform you choose.