Zoom Logo

Vertical Solutions/TSC joint meeting - Shared screen with speaker view
David C Martin
26:54
What about privacy?
David C Martin
27:08
WRT “context sensitive” components
Jim White
28:51
David and everyone - Allan's talk will be about 30 minutes and we'll open up to questions at the end. Those that want to ask, please feel free to unmute and ask your question. If you'd like, ask your question here and I'll bring it up during the Q&A.
Sreyams Jain
29:53
will there be the recording distributed later?
David C Martin
29:54
@Jim White - that was my expectation
Jim White
30:24
@Sreyams - yes, the recording will be made available.
David C Martin
31:48
2) IMHO “long-lived” is a misnomer based on prior methods and apparatus for O/T devices; part of the transformation is applying I/T to O/T; do you see EXF in that role?
Jim St. Leger
33:18
Is EdgeX the only framework Accenture uses and supports? If not, how do they position EdgeX versus other options to customers? What other options? What are advantages of those other options? What would they like to see EdgeX add next as a key feature or capability?
David C Martin
34:40
3) Where is the complex-event-processor or is that relegated to a “Supporting Service”
David C Martin
38:40
4) Do you perform “long-tail” analysis (i.e. historical priors) in the cloud and promulgate back to the edge or is that left to the reader?
Jim St. Leger
41:28
What work or investment does Accenture have in the EdgeX community? Do they have resources contributing directly?
David C Martin
43:22
COMMENT: Highly recommend all network devices may be managed and corresponding agents to collect from same as well as test higher level stack communications
Jim St. Leger
43:55
Have you looked at SDO? Would it solve your certification scheme needs? (asking as I do not know myself.)
Jim St. Leger
44:28
What video analytics solutions have you looked at or are you looking at? Have you looked at OpenVINO?
David C Martin
46:58
WRT privacy more concerned about business data for commercial purposes
David C Martin
47:40
For example: differential privacy, federated learning, ..
Riaz Zolfonoon
48:35
I joined late. Are slides available?
Henry Lau
49:10
Yes, Riaz, slides and recording available on EdgeX Vertical Solutions wiki site afterwards
Jim White
49:17
Yes - the slides and the recording will be made available to all. Stay tuned to the Wiki later today.
Henry Lau
49:27
https://wiki.edgexfoundry.org/display/FA/Vertical+Solutions+Working+Group
Rodney Hess
51:16
Love that Thunderbirds reference!
David C Martin
51:38
RE: Thunderbirds; it’s on again; funded by the South Park guys
David C Martin
52:26
Any suggestions on such gateways? ADLINK w/ the latest nVidia Xavier SOC?
PK
52:43
Referring to AIP+ Edge Agent - do you use HW accelerators at the edge to run the different analytics models? Or is it done in the AIP+ Cloud?
David C Martin
52:56
Do you see K8/K3 need for gateway for HA?
David C Martin
53:35
nVidia, Intel, Google? Any favorites?
David C Martin
54:18
LOL, yes; both; and controllable remotely!
David C Martin
54:28
WRT metrics collection
David C Martin
54:56
Please ask about HA for gateway
Jim White
55:04
will do David.
David C Martin
55:07
High Availability
TimHuckaby
55:08
follow up to PK's question: what "runtime" are you using to execute the ml models at the edge. you mentioned execution on both the gpu and cpu. openvino? onnx runtime? etc....
David C Martin
55:50
@TimHuckaby - containers abstract most of that; the real question is the configuration of the HW and the container runtime
David C Martin
56:29
I see a need for a K3 cluster running Kubeflow, for example
David C Martin
56:57
LOL
David C Martin
57:56
FWIW: github.com/dcmartin/motion-ai for a home-monitoring edge AI system using Home Assistant and Open Horizon
David C Martin
58:11
Crawling toward EXF
PK
58:28
Regarding @TimHuckaby follow-up: a runtime is still relevant regardless if it is running in a container or not
David C Martin
01:00:35
@PK do you mean w.r.t. the model management aspect? I have found that characterizing the device at the HW/OS/Library/Docker level provides the necessary information to determine if the device can execute; the appropriate version of container can be deployed (e.g. aarch64/ubuntu18/cuda10.2/DockerX/runtime=nviai)
David C Martin
01:00:43
runtime=nvidia
David C Martin
01:02:20
It’s the future of computing; pendulum swing back to on-premises versus in-cloud
David C Martin
01:03:46
1) reliability; 2) maintainability; 3) accuracy
Jim White
01:05:24
All - I have captured the questions from this chat window. I'll work with Allan to try to get answers and post those to the Wiki. Thanks to everyone for your participation today. I hope you enjoyed it!
TimHuckaby
01:05:44
@PK, thanks. the question is very relevant. and the speaker...who is awesome...mentioned OV..and some others. in terms of computer vision exectuion at the computer edge, OV does have some competition...again in execution at the edge . and i was curious if Accenture is using any of them.
Jim White
01:06:26
Thundersoft will present on July 14th and Jaingxing Intelligence on July 28th. Please come back for those presentations.
Henry Lau
01:07:44
Check wiki site for upcoming events: https://wiki.edgexfoundry.org/display/FA/Vertical+Solutions+Working+Group
tonyespy
01:10:04
Thanks Allan!
Keith Steele
01:10:25
Thanks Allan great talk
Andrew Foster
01:10:42
Thanks Allan, great talk
Danny
01:10:55
Thank you!
TimHuckaby
01:11:26
best one yet. thank you
wt-nick
01:11:44
thanks