Megh Computing’s PK Gupta joins the conversation to talk about video analytics deployment, personalization, and more.
Megh Computing is a fully customizable cross-platform video analytics solution provider for actionable information in real time. The company was founded in 2017 and is headquartered in Portland, Oregon, with development offices in Bangalore, India.
Co-founder and CEO PK Gupta joins the conversation to talk about delivering analytics, personalization, and more.
As technology continually moves to the edge with video analytics and smart sensors, what are the trade-offs when it comes to implementing the cloud?
GUPTA: The demand for edge analytics is increasing rapidly with the explosion of streaming data from sensors, cameras and other sources. Of these, video remains the dominant data source with over a billion cameras distributed globally. Companies want to extract intelligence from these data streams by using analytics to create business value.
Most of this processing is increasingly done at the edge near the data source. Moving data to the cloud for processing incurs transmission costs, potentially increases security risks, and introduces latencies in response times. Hence the intelligent video analysis [IVA] it’s moving to the limit.
Many end users are concerned about sending video data offsite; what options are available for on-premises computing leveraging the benefits of the cloud?
GUPTA: Many IVA solutions force users to choose between deploying their solution on-premise at the edge or hosted in the cloud. Hybrid models enable on-premise deployments to benefit from the scalability and flexibility of cloud computing. In this model, the video processing pipeline is split between on-premises processing and cloud processing.
In a simple implementation, only the metadata is forwarded to the Cloud for archiving and searching. In another implementation, data import and transformation are done at the edge. Only frames with activity are forwarded to the cloud for processing for analysis. This model is a good compromise between balancing latency and cost between edge computing and cloud computing.
Image-based video analytics has historically needed filtering services due to false positives; how does deep learning reduce them?
GUPTA: Traditional VAT attempts have failed to meet business expectations due to limited functionality and poor accuracy. These solutions use image-based video analytics with computer vision processing for object detection and classification. These techniques are prone to errors that require the deployment of filtering services.
Conversely, techniques that use optimized deep learning models trained to detect people or objects paired with analytics libraries for business rules can essentially eliminate false positives. You can create special deep learning models for custom use cases like PPE compliance, collision avoidance, etc.
We often hear “custom use case” with video AI; what does it mean?
GUPTA: Most use cases need to be customized to meet the functional and performance requirements to deliver VAT. The first level of customization universally required includes the ability to configure monitoring zones in the camera’s field of view, set thresholds for analysis, configure alarms, and set the frequency and recipients of notifications. These configuration capabilities must be provided through a dashboard that uses graphical interfaces to allow users to set up the analysis for proper operation.
The second level of customization involves updating the video analytics pipeline with new deep learning models or new analytics libraries to improve performance. The third level includes training and implementing new deep learning models to implement new use cases, such as a model to detect PPE for worker safety or to count inventory items in a retail store.
Smart sensors such as lidar, presence detection, radar, etc. can they be integrated into an analytics platform?
GUPTA: IVA typically processes video data from cameras only and provides detailed information based on image analysis. And sensor data is typically analyzed by separate systems to produce insights from lidar, radar, and other sensors. A human operator is put into the loop to combine results from disparate platforms to reduce false positives for specific use cases like tailgating, employee authentication, etc.
An IVA platform that can capture data from cameras and sensors using the same pipeline and use machine learning-based contextual analytics can provide insights for these and other use cases. The contextual analysis component can be configured with simple rules and thus can learn to improve the rules over time to provide highly accurate and meaningful insights.