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
Define a balanced VM configuration for ProHawk Vision Server
Measure the improvement of CV results using ProHawk Vision Server
Milestone XProtect Smart Client Plugin functional testing
ProHawk High Availability Assessment
Conclusions and References

Measure the improvement of CV results using ProHawk Vision Server

Home > Workload Solutions > Computer Vision > White Papers > Computer Vision Workflow Enhancement – Image Restoration with ProHawk > Measure the improvement of CV results using ProHawk Vision Server

Measure the improvement of CV results using ProHawk Vision Server

  • Poor lighting, bad weather, and other environmental conditions can degrade the ability of CV applications to detect objects and events of interest. The goal of these tests is to compare the results from CV analytics (using Ipsotek) for vehicle detection with and without ProHawk Vision Server restoration services.

    The different test scenarios used were:

    • Baseline
    • Degraded without ProHawk
    • Restored with ProHawk

    Baseline

    We started with video scene in near ideal conditions and used Ipsotek VISuite AI to detect vehicles of multiple types (cars, trucks, buses, vans) and determined the total counts for those classes over a specified time interval.

    This was our baseline that we refer to as Baseline performance.

    Degraded without ProHawk

    Utilizing video editing software to simulate a low light scenario (corresponding to near night time blackout conditions without street lights or vehicle headlights) to produce a very challenging CV situation.

    Again, we used Ipsotek VISuite AI to detect vehicles of multiple types (cars, trucks, buses, vans) and determined the total counts for those classes over a specified time interval.

    Restored with ProHawk

    A ProHawk was configured to restore the degraded video stream. The auto tune option from ProHawk was not used as from extensive tests we found that applying the ProHawk filters manually gave more optimal results and avoided the misclassification of vehicles.

    Again we used Ipsotek VISuite AI to detect vehicles of multiple types (cars, trucks, buses, vans) and determined the total counts for those classes over a specified time interval.

    Results

    The 3 video streams were synchronized, and counters were in place to count the various vehicle types.

    Table 3. Video stream vehicle detection results
    Scene Car Bus Truck Van
    Baseline 492 2 14 4
    Restored with ProHawk 452 (91%) 4 (200%) 1 (7%) 2 (50%)
    Degraded without ProHawk 324 (65%) 0 (0%) 0 (0%) 0 (0%)

    Findings

    • After enabling ProHawk video restoration (on a terribly degraded video), it was possible to detect 91% of cars that passed through the scene.
    • Despite the really dark scene, Ipsotek was still able to detect about 65% of cars but other types of vehicles were not detectable.
    • ProHawk has a large range of filters that can be applied so it may be possible to improve on the detection rate of 91%.