Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

The AI Edge – School/Education Video Security Monitoring Blueprint

Project Technical Lead:  Hechun Zhang, Baidu, Elected 10/21/19


Blueprint species:

Use Case Attributes

Description

Informational

Type

New

The AI Edge

 

Blueprint Family

Integrated Edge Cloud (IEC)

New

 

Use Case

Safety

and

, security, and surveillance

AI for IoT

 

Blueprint proposed Name

IEC Type 5: AI on the edge

The AI Edge

 

Initial POD Cost (capex)

N/A
Leverage Unicycle POD - less than $150K

 

Scale & Type

One per community

up to 4 servers, x86/ Arm server or deep edge class

With nVIDIA Telsa P4/T4 GPUs

 

Applications

Training

1. Small scale IoT data aggregation and machine learning platform

2. Small scale deep learning models

for IoT devices

training for video data

3. Model training for AcumosApps

 

Power and memory restrictions

IoT devices are power and memory constrained.

Less than 10Kw

K8s 1.12.5 or above- Container Orchestration

OS –CentOS 7.0 or above

 

Infrastructure orchestration

Need IoT device ID correlation with deep learning model (the trained ones in binary format)

Docker 1.13.1 or above

 

SDN

Calico

and K8s, and Containers serving as client/server (a database to collect raw data streams) and server/client (a database for trained models in binary format)

container networking, or OVS-DPDK

 

DataplaneVPP

Workload Type

Containers

(Tensoflow, Keras containers)VMs

 

Additional Details

There are mainly 4 pieces:

  1. a database to collect raw data streams
  2. tons of containers with various models on training
  3. a database for trained models in binary format, and predicted result in JSON format
  4. a database for queries records and transactions of database mentioned in 1 and 3.

Acumos and EdgeX interfaces will be connected in this case

 



Attachment: 

View file
nameThe AI Edge Blueprint-V1.2.pdf
height250


Committer

Affiliation

 

Committer

Committer

Company

 Committer Contact Info

Committer Bio

Committer Picture

Self Nominate for PTL (Y/N)

Hechun Zhang

Baidu

zhanghechun@baidu.com

Seasoned MEC and 5G Solution Architect of Baidu, with impressive 9 years experience in ICT industry.

Project Manager of ODCC Edge Computing Group.


Image Added

                       Y

Dechao KongBaidukongdechao@baidu.com


Gang ChenBaiduchengang13@baidu.com


Zhenhua XuBaiduxuzhenhua02@baidu.com


Siyi HeBaiduhesiyi@baidu.com
Didi



Arm



@Yipan DengIntelyipan.deng@intel.com


PSU




N

Proposed PTL: 

Ken Yi

Contributors: 

@soshun@soshun.arai@arm.com

Tina Tsou 

Wenhui Zhang

Committers: 

Ken Yi 

Raviraj MahatmeArmraviraj.mahatme@arm.com


Tiejun ChenVMwaretiejunc@vmware.com

...