Upgrade to Pro — share decks privately, control downloads, hide ads and more …

An introduction to Kerberos.io

An introduction to Kerberos.io

Kerberos.io initiated as a side project, due to inspiration and motivation in the space of video analytics, computer vision and machine learning. Its first focus was video surveillance only, as nowadays burglary or attacks are very common in this world.

Due to this, our first mission is to provide every human being on this planet with a solution, a video platform, to protect its families, friends, homes or anything else which you think is important.

Our second mission is to make this video platform affordable and Open Source (MIT), and develop it in such a way, that it’s using the latest technologies, to create a seamless, never-seen and delicious user experience.

While we moved forward our third mission is to scale, and make Kerberos reach far beyond a traditional video platform. With the rise of Kerberos Enterprise Suite, we now focus on large scale deployments (covering thousands of cameras), and video analytics through machine learning.

Cédric Verstraeten

October 19, 2021
Tweet

More Decks by Cédric Verstraeten

Other Decks in Technology

Transcript

  1. 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
  2. 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]
  3. 02 Kerberos Agent Low cost video surveillance for anyone, anywhere

    5,000,000+ Product downloads 100,000+ Active instances (estimation)
  4. 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
  5. 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
  6. 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
  7. 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
  8. 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
  9. 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
  10. 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
  11. 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
  12. 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. 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. 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
  15. 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
  16. 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
  17. 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
  18. 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
  19. 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
  20. 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
  21. 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?
  22. 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
  23. 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.