Kerberos.io Introduction Video management and analytics
 at scale where and how you want v2.3 · October 2021 Video analytics for everyone · www.kerberos.io

01 Æ in video surveillance industrÉ Æ at Kerberos.i Æ Expertise mainly in 10+ years experience 7 years computer vision and machine learning Cédric Verstraeten Customer solution advisor [email protected]

02 Kerberos Agent Low cost video surveillance for anyone, anywhere 5,000,000+ Product downloads 100,000+ Active instances (estimation)

03 Kerberos Agent Features Deployments Multi cameras Loop recording to SD card Area motion detection Custom actions/scripts Docker Raspberry Pi (SD flash) Buildroot (KiOS) Balena Cloud

03 Problem statement How do you scale to hundreds or thousands of video streams? How to distribute, load balance, self-healing? How and where to persist data? How to enable edge/cloud synchronisation? Build custom apps, integrations? Computer Vision and ML? Single pane of glass with features

04 Kerberos Enterprise Suite Built for scale with flexible storage, custom integrations and extensions
 and a micro-service based architecture. Kubernetes Scalable deployment Kerberos Agents kerberos vault kerberos hub Storage & Analytics Edge/Cloud Hi-res & on-demand Live streaming Monitoring & analysis Powerful management and more

04 Meet our super stars The coolest projects with the coolest collegues

Deep Kubernetes integration Scalability High availability Fail overs Connect your storage provider Swap storage on-the-fly Connect your workflows Integrate apps & machine learning 05 Kerberos Enterprise Suite Open & extensible ecosystem

06 Kerberos Enterprise Suite Deploy and combine in one or more Kubernetes clusters
 at the edge, in the cloud or hybrid Record Kerberos Agents Scale cameras with Kubernetes High availability Resilient STORE Kerberos Vault On-premise and cloud storage Offline capabilities Extend with API’s and Message Brokers analyze Kerberos Hub Camera monitoring and inventory Recording and livestreaming Grouping and fine grained access Tasks and notifications Event detected ..and more Alerts Follow-up tasks

07 Kerberos Enterprise Suite Deploy and combine in one or more Kubernetes clusters
 at the edge, in the cloud or hybrid Record Kerberos Agents Scale cameras with Kubernetes High availability Resilient STORE Kerberos Vault On-premise and cloud storage Offline capabilities Extend with API’s and Message Brokers analyze Kerberos Hub Camera monitoring and inventory Recording and livestreaming Grouping and fine grained access Tasks and notifications Event detected ..and more Alerts Follow-up tasks

08 Kerberos Agents - Factory Scale your Kerberos Agents

09 Kerberos Agents - Factory Features Minimal footprint No encoding = low memory and CPU usage Recording modes Continuous, time-based or motion-based in region Livestreaming MQTT / WEBRTC Camera agnostic ONVIF / RTSP H264

10 Kerberos Enterprise Suite Deploy and combine in one or more Kubernetes clusters
 at the edge, in the cloud or hybrid Record Kerberos Agents Scale cameras with Kubernetes High availability Resilient STORE Kerberos Vault On-premise and cloud storage Offline capabilities Extend with API’s and Message Brokers analyze Kerberos Hub Camera monitoring and inventory Recording and livestreaming Grouping and fine grained access Tasks and notifications Event detected ..and more Alerts Follow-up tasks

11 Kerberos Vault Bring your own storage or build custom apps

12 Kerberos Vault Features Use own storage Cloud-based or on premise Easy management Seamlessly connect to Kerberos Hub No lock-in Change storage providers on-the-fly On demand uploads Store recordings at the edge, forward on demand to cloud Extensible Integrate and develop own real-time apps Deploy anywhere Via Docker or Kubernetes

13 Kerberos Enterprise Suite Deploy and combine in one or more Kubernetes clusters
 at the edge, in the cloud or hybrid Record Kerberos Agents Scale cameras with Kubernetes High availability Resilient STORE Kerberos Vault On-premise and cloud storage Offline capabilities Extend with API’s and Message Brokers analyze Kerberos Hub Camera monitoring and inventory Recording and livestreaming Grouping and fine grained access Tasks and notifications Event detected ..and more Alerts Follow-up tasks

14 Kerberos Hub Central management and analysis for Kerberos Enterprise Suite

15 Kerberos Hub Features Single interface Manage multiple cameras from one place Grouping Use Sites for location-based camera grouping Powerful filtering Search, filter and drill down through recordings Alerts Event-based notifications Single Sign On Integrate your IDP Live streaming On demand SD/HD streaming Extensible Open API, integrate own analytics and apps Deploy anywhere Via Docker or Kubernetes

16 Architecture Flexible and modular R Each component can be (edge, private/public cloudE R Each component exposes R One or more Kubernetes Clusters, namespaces etc. installed anywhere you want
 Swagger API’g

17 Edge deployment --> Cloud storage Example architecture: Simple scenario I at the edge inside a Kubernetes ClusteF I in the cloud, on preferred hyper scaleF I Connect to your storage Agents Storage Kerberos Vault Kerberos Hub Kerberos Vault STORE Kerberos Hub analyze Cloud Edge Kerberos Agent Record Kerberos Agent Record

18 Edge deployment + storage Example architecture: Offline storage & processing P at the edge inside a Kubernetes ClusteH P at the edge
 and in the cloua P Forward recordings on demand Agents Storage Kerberos Vault STORE Kerberos Vault STORE Kerberos Vault STORE Kerberos Hub analyze Cloud Edge Kerberos Agent Record Kerberos Agent Record

Cloud Edge 19 Private cloud deployment Example architecture: On premise installation t All components within
 private networp t Perpetual licence
 and business continuitP t Support by Kerberos.io
 Operations team Kerberos Vault STORE Kerberos Vault STORE Kerberos Vault STORE Kerberos Hub analyze Kerberos Agent Record Kerberos Agent Record

18 Edge processing with ML Running Machine Learning at the edge W at the edge inside a Kubernetes ClusteU W at the edge
 and in the cloui W Forward recordings from edge to cloud, when ML found a interesting prediction@ W Bring your own machine learning model at scale@ W Loadbalance GPU workloads to a GPU Pool using the NVidia Kubernetes Operator Agents Storage Kerberos Vault STORE Kerberos Vault STORE Kerberos Hub analyze Cloud Edge Kerberos Agent Record Kerberos Agent Record

18 Edge processing with ML Running Machine Learning at the edge Recordings made by Kerberos Agents are stored in Kerberos Vault, and queued until its processed by one of the GPU’s in the pool. 1 x RTX 3080, 2xGTX 1660TI, 2xGTX 1660 Kerberos Vault STORE NVIDIA cluster GPU Pool Kerberos Agent Record Kerberos Agent Record Kerberos Agent Record

18 Usecases - Retail Miniso - Store benchmarking Japanese brand opening retail stores in France. Requires insights through data analytics and video analysis. Challenges9 8 How many customers came in a store4 8 How many customers who entered, bought something4 8 What is the effect of promotions?

18 Usecases - Sports Game2Reel - Sport insights Game2Reel is a new business model focussing on the Padel sports. The platform allows you to keep track of your Padel games, and record your best moments. Challenges6 5 While playing you want to see the entire game afterwards, to analyse specific actionsA 5 Automatic event detection of special moments

18 Usecases - Manufacturing Georgia Pacific - Safety Georgia Pacific is active in the coke industry, with many different production sites across the globe. These sites are fully equiped with machinery, and are not the safest places to work at. Challenges4 3 Make sure that employees are not walking in specific areas, when machines are turned on.

Kerberos.io Cédric Verstraeten [email protected] https://kerberos.io Contact us https://github.com/kerberos-io https://twitter.com/kerberosio https://www.reddit.com/r/kerberos_io/ Kerberos Community v2.3 · October 2021 Video analytics for everyone · www.kerberos.io