KODAKOne Stratus Release - Post Licensing

Looking into our Post Licensing Portal (MVP 0)

Dear KODAKOne Fans,

last time, we gave a few information about the overall platform roadmap. Now, I like to go into our first feature development, where we focus on the management of post licensing. Post licensing is always a little hard to show in the right way because we are talking about infringements. But please think different about the infringement from KODAKOne point of view. Let’s assume, you had just not any chance to find the owner of this image and you took this, with the intention to make a licensing when you will find the creator or owner. 

We are representing the creatives, the owner of such content and search for such users in the internet and help them in exactly this situation to become an official user of this license and therefore become a KODAKOne member in our platform.

Understanding Post Licensing Services

To understand the post licensing service, I start with a short process flow, as deep as I am allowed to dive:

1.     KODAKOne content is uploaded into the platform and marked for “post-licensing”. This upload has amazon S3 as final destination (Managed by mapR XD). On the way to S3 the data is processed in several microservices in our big data driven backend services. Some services are using AI to make a best value prediction, which means that we are searching for relevant image KPIs, such as content, quality, purpose on one hand and licensing predictions for post licensing and licensing based on similar known images. This AI models are trained by our amazing own data science team. A second image stream sends size optimized images and the AI generated image classifier data into the crawler systems. 

2.     The crawlers are small dockers with microservices in Kubernetes with automated scale instances controlled by KAFKA (maps-streams) message loads. Each crawler instance has access to our pattern cloud and KODAKOne Diff hashes to allow Point of Interest and Diff hash similarity detections. Each crawler accesses the internet equipped with many different intelligent search functionality, such as NLP, elastic search and classifiers with a certain amount of requested bandwidth.

3.     Hits on websites are send with sentiments and crawler image classifier data back to our platform brain and stored in our inspector database. This database is used by automated processes, pre-case team members and legal case managers to minimize the effort for clients to decide for enforcements or not.

4.     Enforcements are handled as cases and starting the post licensing process in different client stages by our CRM (currently a customised suiteCRM Framework with data separation on legal country). Stage by stage will be more information via generated mails and requested to clarify the post licensing case and to help the potential infringers to solve this situation and the post licensing portal is given to the infringers to vote for their official licensing. 

5.     At post licensing stage the blockchain plays already an important role for licensing and documentation. The infringement will be documented with its case-id and the related website screenshot image hash are legal requirements, the owner and the asset and the possible licenses are added to the blockchain before already. For asset registration and license rules, Hyperledger Fabric and its chain code are parts of our solution.

6.     A generated mail URL is used to get access to the post licensing portal and helps the infringer to understand the process and starts the licensing request for the future, based on the possible license models, stored in the Hyperledger Fabric and its chain code.

7.     After a successful license agreement, the portal starts the payment process and the payment transactions are triggered in the payment ledger damaged by stellar. Our KODAKCoin is already used internally to make the transactions and the bookings on top of our own stellar protocol implementation. The payment part of the KODAKOne token holds all relevant information about the participants and the license deal.

8.     The sync between both ledgers, as of stellar and the Hyperledger is handled by a trusted party concept and an obligatory linking between both with unique KODAKOne transaction IDs and the stellar bridge service. All transactions between the chains and the KODAKOne backend services are KAFKA decoupled and dockerized in kubernetes. 

9.     After the final payment confirmation from KODAKOne the license transfer will be documented in the licensing part of the KODAKCoin Token attributes and the post licensing process has successfully finished.

The crawlers of KODAKOne

Why do we find so many post licensing content in the internet? This is mostly a question of the right search strategy.  The crawlers, we are using are running 24x7 and are able to go automated and autonomous with a set of rules and models, which helps them to be very effective. We asked a lot of questions at the design phase. 

-      Why we should use an unclassified image in an unclassified web space without any knowledge about the possible reasons for the usage?

-      Could we adjust the search performance, if we know a bit more about the content and the website and the person, which decided to use this image in this particular case?

-      Is there a relation between infringement case, website, image and infringer?

-      Can we predict a conversion rate on search results based on unlicensed image usage to shift to an official license holder?

-      Is it possible to grow our crawler systems on demand into different internet spaces on different purposes?

-      Can we map a crawler search result to a client content and bring a valuation KPI for Post Licensing out of our historic data, that helps us in an IP valuation process?

-      Can we setup watch dogs for the monitoring of ending license periods and found infringements for a stabilized license governance?

We were able to design for most of the questions different technology frameworks, which are running as microservices in a highly automated environment driven by content classifier, website classifier and very efficient hashing and feature extraction and recognition methods. 

E.g. an Italian food image is often used in Spanish and Italian travel agencies, food and fashion shops and blogs worldwide. In this case, crawlers will start searching in these domains first. If they find infringements, they mark the sides, give insights about the ontology, content, imprint, region of the countries, the visitors and the communication channels around. These insights trigger other services, which are responsible for the licensing process and the licensing period monitoring. 

All information together is used in different multi-purpose models, trained in LAB environments, creating classifier for different infringer types, content types and conversion types. The classifiers are running in KAFKA topics (maps-streams) and changing the message flow and the related or involved micro services in the system. 

We found this setup as most powerful for our use cases and best practice for later improvements and team specialization. As result 3 engineers are running currently only in this field and using their time to make it more and more easy and intelligent, to search for new and for similar images in different regions of the internet and our own content.

Post Licensing starts first

It sounds may be a little strength to let the infringers become the first KODAKOne clients on the platform, but it was this idea, which gave us the important visibility at the beginning as well.

The Post Licensing Portal should be available as pre-beta, also mentioned as alpha release for some well-known clients in August / September 2018. The next releases are the official crypto release for ticketing and the OVG Arenas late 2018, the beta release for the KODAKOne licensing platform end of the year and the big KODAKOne marketplace release mid of 2019. In the next weeks I will take the time to explain the coming releases and the very important technologies challenges and innovations on the way from zero to now.

Get involved!

For me this is now an ongoing step to finally being able to share more publicly what we are working on at Wenn Digital to build our KODAKOne platform. Nothing of this would be possible without the truly remarkable development teams and the amazing work with all departments and partners that are involved to make this happen.

Because this will be the second of many posts to come, I would like to hear what your thoughts are on our choices and the overall development of the platform. Well, actually any feedback is more than welcome.

Yours,

Volker