Your Browser is Out of Date

Nytro.ai uses technology that works best in other browsers.
For a full experience use one of the browsers below
Computer Vision Workflow Enhancement – Image Restoration with ProHawk
Overview Solution overview Partner overview Architecture Sizing and scaling guidance Use cases Conclusions and References
Conclusions
References

Conclusions

Home > Workload Solutions > Computer Vision > White Papers > Computer Vision Workflow Enhancement – Image Restoration with ProHawk > Conclusions

Conclusions

  • In this technical white paper, we highlighted the value of ProHawk Vision Server and Plugins as enhancements to the core features of our Dell Validated Design for Computer Vision. We validated the ProHawk software suite with the following use cases scenarios:

    • Define a balanced VM configuration for ProHawk Vision Server
    • Measure the improvement of CV analysis using ProHawk Vision Server
    • Functional appraisal of ProHawk Milestone Smart Client Plugin
    • Assess High Availability capabilities of ProHawk

    We demonstrate once again how the Dell Validated Design for Computer Vision is capable of being a flexible and robust platform starting point for hosting a wide variety of VMS, CV, and related workflow enhancements that meet the needs of large and small organizations from a wide variety of public sector and commercial use cases.

    Define a balanced VM configuration for ProHawk Vision Server

    ProHawk Vision is powered by the parallel processing capabilities of NVIDIA GPU-accelerated technologies to fix dark, obscured, or unclear video. Our development of the Dell Validated Design for Computer Vision has demonstrated the need to carefully balance the mixture of vCPUs, memory, GPU resources, network bandwidth and storage latency to achieve an efficient resource utilization across our VxRail clusters.

    In this validation we compared the difference in performance between designs using NVIDIA A16 and A40 GPUs. These products vary significantly in the balance of hardware-assisted video encoding and decoding resources vs "CUDA cores" typically used for processing CV algorithms.

    We found that the 1/4 of an A16 (A2) GPU paired with 8 vCPUs and 16 GB of memory produced a good tradeoff of performance and manageability. The A16 GPU is compatible with NVIDIA Enterprise vGPU software that integrates with VMware to add many operational benefits over traditional "pass-through" mode GPU configurations in virtualized environments.

    Measure the improvement of CV analysis using ProHawk Vision Server

    We successfully measured the improvement of CV results using ProHawk Vision Server by comparing the following types of scenes:

    • Baseline
    • Degraded without ProHawk
    • Restored with ProHawk

    We demonstrated how ProHawk Vision Server can turn dark, obscured, or unclear live video into sharper and clearer video that improved the detection rate of cars by at least 26%.

    Functional appraisal of ProHawk Smart Client Plugin

    The ProHawk Smart Client Plugin is designed to directly integrate with the Milestone XProtect Smart Client. ProHawk running as a plug-in to the Milestone XProtect Smart Client enables users to adjust computer vision restoration settings in real-time, zoom in and pan on restored video from within a Milestone XProtect Smart Client.

    We found the installation and configuration of the Milestone XProtect Smart Client plug-in to be intuitive, consistent , and easy to configure. Integration with Milestone XProtect was straightforward.

    • ProHawk Plugin offers additional functionality to the Milestone XProtect Smart Client UI
    • Several different camera steams can be place within XProtect UI tile views.
    • ProHawk client requires access to GPU resources.

    Assess High Availability capabilities of ProHawk

    High Availability is an important aspect of the Dell Validated Design for Computer Vision. During our testing we identified that no native HA capabilities exist the ProHawk Vision servers. This can be mitigated somewhat by recording the raw video stream to ensure no video loss if a ProHawk Vision server fails.

    The Dell Validated Design for Computer Vision also has infrastructure HA by default. This means that if a VM has a failure it will be reinstantiated in another part of the cluster.