With NVIDIA Cosmos Curator, Milestone Announces Platform to Enable Access to Data and Train AI Models

  • Mar 19, 2025

AI developers across the globe share the challenge of finding sufficient high-quality, accurate and reliable video data to train visual AI models. Today, Milestone Systems announces Project Hafnia to build services and democratize AI-model training with high-quality, compliant video data leveraging NVIDIA Cosmos Curator and AI model, fine-tuning microservices.

Leveraging the NVIDIA platform, Milestone Systems’ Project Hafnia is aiming to be a leading service for data generators, to share and utilize their data, and for developers, to access traceable and regulatory-compliant annotated video data.

Milestone’s goal is to put the latest advances in Vision Language Models (VLM) and supporting data curation capabilities in the hands of as many developers as possible. One of the first service offerings is a VLM, fine-tuned using NVIDIA fine-tuning microservices on a large volume of compliant transportation data, curated using NVIDIA Cosmos Curator.

The VLM’s accuracy and performance optimizations are tuned for running on NVIDIA GPUs and in NVIDIA’s video search and summarization (VSS) AI blueprint.

“By leveraging the NVIDIA platform, Milestone Systems is helping accelerate this next wave of powerful visual services,” says Deepu Talla NVIDIA Vice President and General Manager — Embedded and Edge Computing. “The next phase in development and adoption of visually perceptive Agentic AI services will be unlocked by recipes like NVIDIA VSS blueprint combined with widely available and accessible fine-tuned VLM models,” he says. Milestone Systems, a global leader in data-driven video technology and analytics, is aggregating compliant data through its global network of partners and customers in video data management. These organizations seek to leverage their own data to develop smarter analytics and will be able to benefit from the platform, which can be utilized to train AI models on sufficient, compliant data.

“Artificial intelligence is our generation’s biggest game-changer. A major challenge for the ongoing development is having access to enough high-quality data for training AI models. The Project Hafnia platform will collect and curate data with the aspiration to be the world's smartest, fastest and responsible platform for video data and training of AI models,” says Thomas Jensen, CEO of Milestone Systems.

Additionally, Project Hafnia will remove major friction points and provide smooth and seamless access to data. Driven by NVIDIA Cosmos Curator data curation, the service will speed up developing AI and analytics – up to 30 times faster than today’s standards.

Milestone Systems will offer two services:

A cutting-edge service for AI model training with high-quality video data through Training as a Service, where software developers can access quality data to train AI-models

A new Visual Language Model as a Service for smart city transportation and Intelligent Traffic Service use-cases that will be industry leading in performance.

With new high-quality data and improved annotation, future analytic software for applications such as traffic management, manufacturing, airports, law enforcement, and business could achieve accuracy high enough to enable large-scale automation of operations and surveillance like never before.

By leveraging the NVIDIA platform, Milestone Systems has taken the first step in developing a market leading traffic and transportation Visual Language Model (VLM). The model supports a range of use cases, including general traffic assessments, driving condition evaluations, alert validation, and incident reporting.

Project Hafnia is initially launching as a pilot, with developers having the opportunity to sign up for early access at https://hafnia.milestonesys.com/joinwaitlist. The platform will initially focus on video data from traffic with plans to broaden the scope once fully live. Milestone Systems owns and manages the platform.

Featured

  • Evolving Cybersecurity Strategies: Uniting Human Risk Management and Security Awareness Training

    Organizations are increasingly turning their attention to human-focused security approaches, as two out of three (68%) cybersecurity incidents involve people. Threat actors are shifting from targeting networks and systems to hacking humans via social engineering methods, living off human errors as their most prevalent attack vector. Whether manipulated or not, human cyber behavior is leveraged to gain backdoor access into systems. This mainly results from a lack of employee training and awareness about evolving attack techniques employed by malign actors. Read Now

  • Report: 1 in 3 Easily Exploitable Vulnerabilities Found on Cloud Assets

    CyCognito recently released new research highlighting critical security vulnerabilities across cloud-hosted assets, revealing that one in three easily exploitable vulnerabilities or misconfigurations are found on cloud assets. As organizations increasingly shift to multi-cloud strategies, the findings underscore significant security gaps that could provide attackers with potential footholds into networks. Read Now

  • Built for Today, Ready for Tomorrow

    Selecting the right VMS is critical for any organization that depends on video surveillance to ensure safety, security and operational efficiency. While many organizations focus on immediate needs such as budget and deployment size, let us review some of the long-term considerations that can significantly impact a VMS's utility and flexibility. Read Now

  • Paving the Way to Smart Buildings

    In today's rapidly evolving security landscape, the convergence of on-prem, edge and cloud technologies are critical. The physical security landscape is undergoing a profound transformation, driven by the rapid digitalization of buildings and the evolving needs of modern organizations. As the buildings sector pivots towards smart, AI and data-driven operations, the integration of both edge and cloud technology has become crucial. Read Now

  • The Cybersecurity Time Bomb

    If you work in physical security, you have probably seen it: a camera, access control system, or intrusion detection device installed years ago, humming along without a single update. It is a common scenario that security professionals have come to accept as "normal." But here is the reality: this mindset is actively putting organizations at risk. Read Now

Most   Popular

New Products

  • Unified VMS

    AxxonSoft introduces version 2.0 of the Axxon One VMS. The new release features integrations with various physical security systems, making Axxon One a unified VMS. Other enhancements include new AI video analytics and intelligent search functions, hardened cybersecurity, usability and performance improvements, and expanded cloud capabilities

  • Mobile Safe Shield

    Mobile Safe Shield

    SafeWood Designs, Inc., a manufacturer of patented bullet resistant products, is excited to announce the launch of the Mobile Safe Shield. The Mobile Safe Shield is a moveable bullet resistant shield that provides protection in the event of an assailant and supplies cover in the event of an active shooter. With a heavy-duty steel frame, quality castor wheels, and bullet resistant core, the Mobile Safe Shield is a perfect addition to any guard station, security desks, courthouses, police stations, schools, office spaces and more. The Mobile Safe Shield is incredibly customizable. Bullet resistant materials are available in UL 752 Levels 1 through 8 and include glass, white board, tack board, veneer, and plastic laminate. Flexibility in bullet resistant materials allows for the Mobile Safe Shield to blend more with current interior décor for a seamless design aesthetic. Optional custom paint colors are also available for the steel frame.

  • AC Nio

    AC Nio

    Aiphone, a leading international manufacturer of intercom, access control, and emergency communication products, has introduced the AC Nio, its access control management software, an important addition to its new line of access control solutions.