SensiML Onjezani Kukonzekera Kwambiri mu Malangizo a Zida Zomangamanga

Chiyambi cha Host - mphindi 5
Yambitsani malingaliro ndi cholinga cha lab - mphindi 10
Kukonzekera kwa "Real-time" kwa ndondomeko ya sitepe ndi sitepe yopanga zitsanzo - mphindi 60
- Flash SensiML yogwirizana ndi firmware yosonkhanitsira deta ku Thunderboard Sense 2 (TBS2)
- Konzani ndikulumikiza TBS2 ku SensiML Data Capture Lab
- Jambulani data ya 'slide demo' ndi bolodi yopanda kanthu (ogwiritsa sadzakhala ndi zida za Fan)
- Lembani deta ndi kusunga ndi sample project (sitidzagwiritsa ntchito nthawi yotsalayo)
- Invoke Analytics Studio (panthawiyi, ogwiritsa ntchito azigwira ntchito kuchokera ku dataset yotsatiridwa ya TBS2)
- Gwirani ntchito ndi masitepe omanga mafani amtundu wa fan state
- Pangani Paketi Yachidziwitso
- Zosankha: Mtundu wa Flash kupita ku TBS2

Kanema wachiwonetsero wa Smart Building Applications - mphindi 5
Q&A - mphindi 10

Chiyambi cha SensiML

SensiML ndi kampani ya B2B yopanga mapulogalamu a AI pamphepete mwa IoT
- Imathandiza omanga kupanga ma ultra-compact ML sensors opanda ukatswiri wa sayansi ya data
- Zitsanzo zazing'ono ngati 10KB!
- Gulu lakale la Intel Curie/Quark MCU AI la zida zamapulogalamu, latsala kuti lipange SensiML mu 2017.
Silicon Labs ndi SensiML Solution
- Kubweretsa mphamvu ya ML ku banja la EFR32/EFM32 MCU
- Rapid smart IoT application prototyping yokhala ndi Thunderboard Sense 2

SensiML ili ndi kukhazikika komanso chithandizo chapadziko lonse lapansi
- Adapezedwa mu 2019 ndi QuickLogic Corp; khazikitsani ndikuyendetsa ngati pulogalamu yodziyimira payokha (yochokera ku Portland, OR)
- Okhazikitsa ma tchanelo (Avnet, Future Electronics, Mouser, Shinko Shoji)
- Maofesi ogulitsa / othandizira ku UK, US, Japan, Taiwan, China

Mwayi Wa TinyML mu Smart Buildings

Zovuta Zomwe Zilipo za Smart IoT Sensor Application Development

Cloud-Centric AI

High Network Traffic Load
Kukwera kwakukulu
Osalekerera Zolakwa
Chiwopsezo Chosadziwika cha Chitetezo cha Data
Nkhawa Zazinsinsi

Kuphunzira Mwakuya

Zofunikira zazikulu za data yophunzitsira
Chidziwitso chachikulu cha kukumbukira
High processing ntchito
Kugwiritsa ntchito mphamvu kwambiri
Moyo wa batri woyipa womaliza

Mapeto Olembedwa Pamanja

Pang'onopang'ono komanso wogwira ntchito molimbika
Kukula kwa code kosadziwika patsogolo
Katswiri wa sayansi ya data yochepa
Ma library ovuta a AI/ML
Osasinthika / opikisana

TinyML = IoT Edge ML + AutoML

IoT Edge ML: Mapeto a Autonomous
- Mayendedwe ang'onoang'ono a netiweki komanso moyo wautali wa batri wopanda zingwe
- Palibe kukonza mtambo kapena kudalira maukonde
- Kuyankha kwenikweni
AutoML: Konzani Popanda Katswiri wa AI
- Auto-optimizer imasankha mtundu wabwino kwambiri wazomwe zaperekedwa
- Kuphunzira kwamakina achikale (ML) kudzera pakuphunzira mozama
- SensiML TinyML imatulutsa mitundu yaying'ono ngati 10KB!

Kulemba pamanja sikufunikira
- Khodi yachitsanzo yopangidwa yokha kuchokera kumagulu a maphunziro a ML
- Imapulumutsa miyezi yambiri yachitukuko, komanso ukatswiri wa sayansi ya data
- Wopanga mapulogalamu amatha kusintha gawo lililonse la code ya AutoML momwe amafunira

Ntchito Yomanga Model

Jambulani Data

Nthawi: Maola mpaka Masabata* (Kutengera zovuta zosonkhanitsira deta)
Luso: Katswiri wa Domain (Monga zimafunikira kusonkhanitsa ndikulemba zochitika zomwe zimakonda)

Zindikirani: Tigwiritsa ntchito zina zomwe zidasonkhanitsidwa kale kuti tifulumizitse sitepe iyi ya msonkhano

Mangani Chitsanzo

Nthawi: Mphindi mpaka Maola (Malingana ndi kuchuluka kwa chiwongolero choperekedwa)

Luso: Palibe (Full AutoML)
- Basic ML Concepts (Advanced UI tuning)
- Python Programming (Kuwongolera kwathunthu mapaipi)

Yesani Chipangizo

Nthawi: Mphindi mpaka Masabata (Malingana ndi zosowa zophatikizira mapulogalamu)

Luso: Palibe (Binary firmware yokhala ndi code yodzipangira yokha ya I/O)
Embedding Programming (Kuphatikizika kwa library ya SensiML kapena gwero la C ndi code ya ogwiritsa)

Zolinga za Msonkhano

Tsegulani zida za SensiML za TinyML ndi njira yomangira yachitsanzo pa Silicon Labs Thunderboard Sense 2

Dziwani zambiri zoyendetsedwa ndi data kuyang'aniridwa ndi ML sensor algorithm development
Phunzirani kayendetsedwe ka ntchito kuchokera pakusonkhanitsira deta kudzera mukutsimikizira ndi kuyesa pazida zomangira mitundu ya IoT
Pangani mtundu wogwira ntchito wa HVAC wolosera woyambira mpaka kumapeto
Yankhani mafunso omwe mungakhale nawo okhudza njira yopangira machitsanzo a TinyML

Ntchito Yogwira Ntchito ya HVAC Predictive Maintenance Application

Zolinga za gawo lathu lothandizira, tipanga chida chanzeru chowunikira mafani
Mafani omwe amagwiritsidwa ntchito ponseponse pomanga makina a HVAC: Zowombera, kuziziritsa kwa zida, zowongolera mpweya, ma ducting mpweya.

Kulephera kapena kuwonongeka kungayambitse kuchepa kwa mphamvu, kuchuluka kwa mphamvu zamagetsi, kulephera kwa HVAC
Tipanga chida chosavuta chowunikira chomwe chimatha kuzindikira mafani angapo abwinobwino komanso osadziwika bwino:
- Kuzimitsa / kuyatsa
- Zokwera zomasuka
- Kutchinga kwa mafani
- Mpweya wotsekedwa pang'ono kapena wotsekedwa kwathunthu
- Kuwonongeka kwa blade
- Kugwedezeka kwakukulu

Tiyeni tiyambe ndondomekoyi

"Real-time" workshop sitepe ndi sitepe yopangira chitsanzo - mphindi 60

Flash SensiML yogwirizana ndi firmware yosonkhanitsira deta ku Thunderboard Sense 2 (TBS2)
Konzani ndikulumikiza TBS2 ku SensiML Data Capture Lab
Jambulani data ya 'slide demo' ndi bolodi yopanda kanthu (ogwiritsa sadzakhala ndi zida za Fan)
Lembani deta ndi kusunga ndi sample project (sitidzagwiritsa ntchito nthawi yotsalayo)
Invoke Analytics Studio (panthawiyi, ogwiritsa ntchito azigwira ntchito kuchokera pagulu lachiwonetsero la TBS2 lomwe lasonkhanitsidwa kale)
Gwirani ntchito ndi masitepe omanga mafani amtundu wa fan state
Pangani Paketi Yachidziwitso
Mtundu wa Flash ku TBS2

Kanema wa Demo

Zolemba / Zothandizira

SensiML Onjezani Kukonzekera Kwambiri mu Zida Zomangamanga Zanzeru [pdf] Malangizo
Onjezani Predictive Maintenance mu Smart Building Devices, Kukonza mu Smart Building Devices, Smart Building Devices, Building Devices

Maumboni

Buku Logwiritsa Ntchito

SensiML Onjezani Kukonzekera Kwambiri mu Zida Zomangamanga Zanzeru

Agenda