• D1.2 Pro­ject hand­book with stra­tegic plan for qual­ity assur­ance and risk man­age­ment (M2)
  • D1.3 First Year Annu­al Report (M12)
  • D1.4 Second Year Annu­al Report (M24)
  • D7.1 Iden­ti­fic­a­tion and recruit­ment cri­ter­ia (M6)
  • D7.2 Informed con­sent pro­ced­ures (M6)
  • D7.3 Writ­ten Inform­a­tion Sheet and Informed Con­sent Form (M6)
  • D7.4 DPO appoint­ment and data pro­tec­tion policy (M6)
  • D7.5 Data trans­fer from EU to Korea (M6)
  • D7.6 Data trans­fer from Korea to EU (M6)
  • D7.7 Social impact assess­ment (M6)
  • D7.8 Eth­ics Advisor appoint­ment (M6)
  • GENERALIZED PRODUCT QUANTIZATION NETWORK FOR SEMI-SUPERVISED IMAGE RETRIEVAL: Young Kyun Jang, Nam Ik Cho, Com­puter Vis­ion and Pat­tern Recog­ni­tion (CVPR); June 2020
  • A PLATFORM PROVIDING INTERACTIVE SIGNAGE BASED ON EDGE-CLOUD COOPERATION: Jae­won Moon, Seung­woo Kum, Sang­won Lee; Korean Journ­al of Inter­net Com­put­ing and Ser­vices, Journ­al of Inter­net Com­put­ing and Ser­vices (KRO), vol.20, no.2, pp. 39–49, 2019
  • AI MANAGEMENT PLATFORM WITH EMBEDDED EDGE CLUSTER: Seung­woo Kum, Mis­eon Yu, Young­kee Kim, Jae­won Moon and Silvio CretiICCE 2021 (39th IEEE Inter­na­tion­al Con­fer­ence on Con­sumer Elec­tron­ics); 10–12 Janu­ary 2021, Online Event
  • DESIGN AND IMPLEMENTATION OF OBJECT DETECTOR BASED ON IOMT STANDARD: Seung­woo Kum, Jae­won Moon, Young­kee Kim; KIBME (Korean Insti­tute of Broad­cast­ing and Media Engin­eers) 2019 Sum­mer Con­fer­ence (KR)
  • DEVELOPMENT OF DOOR OPEN-CLOSE PREDICTION MODEL BASED ON INDOOR AIR DATA AND LSTM MODEL: Sangyeon Cho, Jae­won Moon, Seung­woo Kum; KSII (Korean Soci­ety for Inter­net Inform­a­tion) 2019 Spring Con­fer­ence
  • DEEP CLUSTERING AND BLOCK HASHING NETWORK FOR FACE IMAGE RETRIEVAL: Young Kyun Jang, Dong-ju Jeong, Seok Hee Lee, and Nam Ik Cho; 14th Asi­an Con­fer­ence on Com­puter Vis­ion (ACCV), 2018
  • LEARNING BACKGROUND SUBTRACTION BY VIDEO SYNTHESIS AND MULTI-SCALE RECURRENT NETWORKS: Sun­gk­won Choo, Wonkyo Seo, Dong-ju Jeong, and Nam Ik Cho; 14th Asi­an Con­fer­ence on Com­puter Vis­ion (ACCV), 2018
  • NOVEL METHOD FOR NEURAL NETWORK ON EDGE DEVICE: Seung­woo Kum, Jae­won Moon, Young­kee Kim; IEMEK (Insti­tute of Embed­ded Engin­eer­ing of Korea) 2018 Sum­mer Con­fer­ence (KR)
  • CLOUD-EDGE COLLABORATION FRAMEWORK FOR IOT DATA ANALYTICS: Jae­won Moon, Saungyeon Cho, Seung­woo Kum, Sang­won Lee; Inter­na­tion­al Con­fer­ence on ICT Con­ver­gence 2018 (Korean Insti­tute of Com­mu­nic­a­tion and Inform­a­tion Sci­ence)
  • TUDI GRADITI JE MOGOČE PAMETNO (eng.: IT IS POSSIBLE TO BUILD SMART): Janez Brežnik, Uroš Pašč­in­ski, Vlado Stankovski; News­pa­per DELO (SLO), 2019, May 16, Vol.61/no.110, p.14 (prin­ted ver­sion)
Benefits of the EU-KR collaboration and lessons learnt

The Hori­zon 2020 joint EU-KR pro­gram gran­ted the involved par­ti­cipants and their eco­sys­tems a frame­work to estab­lish a cooper­a­tion that would not have taken place oth­er­wise. While EU insti­tu­tions are usu­ally cooper­at­ing with oth­er part­ners in the con­tin­ent, Korean research­ers do not have many chances to carry out inter­na­tion­al R&D pro­jects. They usu­ally focus on domest­ic pro­jects, along with a few inter­na­tion­al pro­grams such as Eureka or Eurostar. Though con­sor­tia in these pro­grams con­sist of part­ners from vari­ous nations, the inter­ac­tion between par­ti­cip­at­ing insti­tu­tions are rather lim­ited when com­pared to the vibrant and act­ive col­lab­or­a­tion estab­lished with­in the Hori­zon 2020 joint EU-KR pro­gram. Beside the per­son­al exper­i­ence and growth, which per se con­sti­tute an immense value (a per­son with wider and more inclus­ive per­spect­ive is an enrich­ment for the soci­ety), the pro­gram gave research­ers from both sides the oppor­tun­ity to inter­act with soci­et­al and indus­tri­al eco­sys­tems that are quite dif­fer­ent from each oth­er, where customers/consumers and gen­er­al pub­lic have diverse needs or pref­er­ences, some­times ori­ent­ing the devel­op­ment of innov­at­ive tech­no­lo­gies towards het­ero­gen­eous dir­ec­tions. Each of the research­ers involved was able to con­front with vari­ous tech­nic­al per­spect­ives and reg­u­la­tions, and con­trib­ute, in col­lab­or­a­tion with its peers, to inter­na­tion­al stand­ards and com­munit­ies.

Technical viewpoint

DECENTER widened our under­stand­ing of indus­tri­al and aca­dem­ic devel­op­ments related to cloud com­put­ing, IoT and AI tech­no­lo­gies, which are very dif­fer­ent by nature, yet are matur­ing very quickly. Indeed, cloud and IoT tech­no­lo­gies revolu­tion­ised the way a ser­vice is provided, with Dock­er and Kuber­netes, two of the tech­no­lo­gies lever­aged by DECENTER, already de-facto stand­ards solu­tions for ser­vice deploy­ment after less than ten years from their first appear­ance. With respect to AI, research dir­ec­tions are man­i­fold (e.g., increas­ing accur­acy with dif­fer­ent archi­tec­tures, data manip­u­la­tion, optim­isa­tion, etc.). Though the top­ics them­selves are quite rel­ev­ant and inter­est­ing to the research­ers, it takes the effort of dif­fer­ent experts to identi­fy which of them are of import­ance to dif­fer­ent groups of stake­hold­ers to real­ise suc­cess­ful AI ser­vices.

This joint EU-KR pro­gram – more spe­cific­ally, the DECENTER pro­ject – provided a chance to organ­ize a con­sor­ti­um with dif­fer­ent back­grounds to deal with the chal­lenges stated in the call. At the time of the pro­pos­al pre­par­a­tion, EU part­ners con­trib­uted with sol­id know­ledge on cloud, edge and IoT tech­no­lo­gies (with more focus on the plat­form side), while Korean part­ners had a strong expert­ise on Arti­fi­cial Intel­li­gence (more focus on the tools and ser­vices side). From the tech­nic­al stand­point, this was the oppor­tun­ity to mix up know­ledge start­ing from com­mon interests and pur­sue ambi­tious object­ives, grant­ing a tech­no­logy step up and the uptake of rel­ev­ant out­comes. Part­ners from EU and KR were col­lab­or­at­ing on the pro­ject use cases, exploit­ing the oppor­tun­ity to work on many, indus­tri­ally rel­ev­ant, scen­ari­os. This is, for instance, the case of the robot­ic logist­ics (pro­posed by a European SME) or the ambi­ent intel­li­gence (pro­posed by a Korean research centre) use cases, in which both con­trib­uted know­ledge and assets. A rel­ev­ant amount of expert know­ledge was trans­ferred, e.g., cloud orches­tra­tion from EU to KR and AI mod­el optim­isa­tion the oth­er way around; to facil­it­ate cross-pol­lin­a­tion, sem­inars and meet­ings were organ­ised through­out the pro­ject dur­a­tion.

Regulatory viewpoint

The joint effort star­ted shortly after the intro­duc­tion of the Gen­er­al Data Pro­tec­tion Reg­u­la­tion (GDPR). The Korean equi­val­ent of GDPR, the Private Inform­a­tion Pro­tec­tion Act (PIPA) was announced dur­ing the second year of the pro­ject. Besides the lack of inform­a­tion that both sides had of counterpart’s reg­u­la­tion, research car­ried out by con­sor­ti­um mem­bers did not include any pri­vacy meas­ure until recently. While imple­ment­ing the ser­vices on the DECENTER plat­form, we were able to identi­fy reg­u­la­tions on both sides and learned how to imple­ment them with the pro­posed tech­no­lo­gies. It was stun­ning that we could dir­ectly imple­ment some of them with our block­chain-based frame­work when shar­ing AI mod­els across bor­ders.

In Korea there are a few organ­iz­a­tions that help com­pan­ies to meet the require­ments on dif­fer­ent reg­u­la­tions of tar­get mar­kets (e.g., the GDPR cen­ter of KISA), but even though issues can be iden­ti­fied, it takes a con­sid­er­able amount of time to adapt tech­no­lo­gies to be com­pli­ant with GDPR. On the oth­er hand, there is very little or no know­ledge of PIPA reg­u­la­tion in Europe. DECENTER exper­i­ence is a valu­able asset for EU and Korean part­ners to make a softer land­ing on the respect­ive mar­kets and be pre­pared to extend their pos­i­tion in oth­ers.

Standardisation and outreach viewpoint

In the inter­na­tion­al stand­ard­isa­tion bod­ies addressed by DECENTER, con­trib­ut­ors rep­res­ent their nation; in oth­er com­munit­ies and fora, con­trib­ut­ors rep­res­ent the com­pany or the organ­iz­a­tion they belong. Fur­ther, stand­ard­isa­tion is dir­ectly con­nec­ted to Intel­lec­tu­al Prop­erty Rights (IPR) and can be used to pro­tect it, which makes it hard to observe a fruit­ful col­lab­or­a­tion between dif­fer­ent organ­isa­tions.

The EU-KR pro­gram facil­it­ated col­lab­or­a­tion among part­ners. First, the uni­fied tech­nic­al vis­ion required little dis­cus­sion on work items or par­ti­cip­a­tion on stand­ard­isa­tion activ­it­ies. Second, grant and coordin­a­tion agree­ments set a com­mon and pre­cise back­ground for the cooper­a­tion, redu­cing risks of con­flicts on IPR. In the DECENTER pro­ject, the con­sor­ti­um sub­mit­ted three joint con­tri­bu­tions to ISO/IEC JTC1 SC29 WG7, offer­ing a wide vis­ib­il­ity to our work on two rel­ev­ant top­ics – the deploy­ment of AI through cloud tech­no­lo­gies, and the con­fig­ur­a­tion of AI meth­ods to build an end-to-end ser­vice. The lat­ter con­tri­bu­tion has now been selec­ted as a next work item of WG7. In addi­tion to that, DECENTER con­trib­uted to a few domest­ic and inter­na­tion­al com­munit­ies with mem­bers from both EU and KR, such as the Urb­an Tech­no­logy Alli­ance or Korean Insti­tute of Broad­cast and Media. These DECENTER out­reach­ing activ­it­ies attrac­ted interests from research soci­ety – part­ners have received mul­tiple invit­a­tions for events and new research pro­jects. The fact that con­tri­bu­tions included mem­bers from both sides gave cred­ib­il­ity to our work, inspired trust in the eyes of inter­locutors, and eased off­line dis­cus­sions.

There­fore, two sig­ni­fic­ant bene­fits of the EU-KR joint pro­gram on inter­na­tion­al stand­ard­isa­tion were noticed. First, it lowered bar­ri­er for col­lab­or­a­tion, from the tech­nic­al, leg­al and admin­is­trat­ive view­points. Second, co-author­ing of con­tri­bu­tions had a note­worthy effect on attract­ing atten­tion to our work and giv­ing a pos­it­ive impres­sion, rep­res­en­ted a mutu­al con­sensus between organ­isa­tions and coun­tries, pro­mot­ing a uni­fied EU-KR angle on mat­ters that will influ­ence the tech­no­lo­gies at the basis of our digit­al future. 

Stakeholders’ workshop on business scenarios

DECENTER pro­ject has developed some poten­tial busi­ness scen­ari­os for the pro­ject out­comes which were val­id­ated with a set of extern­al stake­hold­ers to the pro­ject. The pro­ject presen­ted the busi­ness mod­els to the stake­hold­ers in two work­shops and then reques­ted for their feed­back to an online sur­vey of 12 ques­tions about impact assess­ment and adop­tion feas­ib­il­ity of pro­ject res­ults.

A total of 21 stake­hold­ers were selec­ted by the pro­ject part­ners with­in their own organ­iz­a­tions but oth­er depart­ments or busi­ness units (7 out of the total) and with­in their con­tacts net­work (14 out of them). They were from both Europe and Korea and from dif­fer­ent pro­files to provide a com­ple­ment­ary view on the pro­ject from diverse angles, thus we involved cloud/edge infra­struc­ture pro­viders, uni­ver­sit­ies and com­pan­ies as tech­no­logy innov­at­ors, sys­tem integ­rat­ors and end users.

The con­sulta­tion was a very use­ful instru­ment to assess the assump­tions about pos­sible impact and adop­tion of DECENTER in poten­tial cus­tom­ers. It has showed the rel­ev­ance of DECENTER res­ults in the mar­ket, the value it brings to the busi­ness and the require­ments are needed to set up DECENTER in a real scen­ario. Thanks to the obtained feed­back, some improve­ments have been achieved dur­ing the remain­ing pro­ject and oth­ers are iden­ti­fied for the next future.