30 Aug 2024
|
Mike Boland
AWE Talks: Building a Decentralized Visual Positioning System
AWE USA 2024
Welcome back to AWE Talks, our series that revisits the best of AWE’s conference sessions. With AWE USA 2024 recently concluded, we have a fresh batch of footage to keep us busy for weeks to come.
We continue the action this week with a look at AR's longstanding promise of elevating retail experiences through consumer wayfinding and merchant-facing utilities. It's a lot harder than it looks...
See the summarized takeaways below, along with the full session video. Stay tuned for more video highlights each week and check out the full library of AWE USA 2024 sessions on AWE’s YouTube Channel.
Speakers
Nils Pihl, Auki Labs
Key Takeaways & Analysis
– One of AR's longstanding promises is indoor wayfinding such as airports & retail stores.
– An early ARKit app demo captured the world's attention with this use case (see it here).
– But there was one catch... it didn't exist. It was just a simulated concept demo.
– This is simply because this use case is immensely challenging to execute in practice.
– Among other things, it requires data on store layouts, product placement, etc.
– This isn't data that Google or anyone else has, and GPS degrades indoors.
– Another challenge is that retailers guard their merchandising strategies closely.
– So they don't share interior store data with centralized entities like Google.
– Many also refuse to store such data on any cloud, given its strategic sensitivity.
– Even if you did have a reliable store digital twin, it would quickly be out of date.
– Digital twins are too static, given shifting in-store layouts and inventory dynamism.
– For all these reasons, a visual positioning system may be better to build spatial maps.
– Computer vision and AI can work together for spatial and semantic understanding.
– But for the above reasons, centralized VPS systems go against retailer sensitivities.
– That includes systems such as Google and Niantic's VPS technologies.
– The answer may instead be a decentralized system of visual positioning.
– Devices can capture visual data while stored and processed in a decentralized way.
– This has the scale advantages of crowdsourcing but without centralized cloud storage.
– The end goal is to utilize billions of devices to collectively assemble spatial maps.
– Store cameras can be enlisted to do the same thing, aligned with retailers' goals.
– Furthermore, the collaboration of many devices engenders greater insight & dimension.
– Practical benefits again include consumer in-store wayfinding and product search.
– There are also merchant use cases like task management (e.g., inventory optimization).
– Being able to quickly restock items could save retailers $1 trillion in annual lost revenue.
Speakers
Nils Pihl, Auki Labs
Key Takeaways & Analysis
– One of AR's longstanding promises is indoor wayfinding such as airports & retail stores.
– An early ARKit app demo captured the world's attention with this use case (see it here).
– But there was one catch... it didn't exist. It was just a simulated concept demo.
– This is simply because this use case is immensely challenging to execute in practice.
– Among other things, it requires data on store layouts, product placement, etc.
– This isn't data that Google or anyone else has, and GPS degrades indoors.
– Another challenge is that retailers guard their merchandising strategies closely.
– So they don't share interior store data with centralized entities like Google.
– Many also refuse to store such data on any cloud, given its strategic sensitivity.
– Even if you did have a reliable store digital twin, it would quickly be out of date.
– Digital twins are too static, given shifting in-store layouts and inventory dynamism.
– For all these reasons, a visual positioning system may be better to build spatial maps.
– Computer vision and AI can work together for spatial and semantic understanding.
– But for the above reasons, centralized VPS systems go against retailer sensitivities.
– That includes systems such as Google and Niantic's VPS technologies.
– The answer may instead be a decentralized system of visual positioning.
– Devices can capture visual data while stored and processed in a decentralized way.
– This has the scale advantages of crowdsourcing but without centralized cloud storage.
– The end goal is to utilize billions of devices to collectively assemble spatial maps.
– Store cameras can be enlisted to do the same thing, aligned with retailers' goals.
– Furthermore, the collaboration of many devices engenders greater insight & dimension.
– Practical benefits again include consumer in-store wayfinding and product search.
– There are also merchant use cases like task management (e.g., inventory optimization).
– Being able to quickly restock items could save retailers $1 trillion in annual lost revenue.
– All the above inherits the virtues of DePIN (decentralized physical infrastructure).
– Think of this as a sort of mashup of computer vision, AI, Web3, and IoT.
– Expect to hear more about DePIN as spatial computing and AI continue to converge.
– Meanwhile, Auki's Convergent platform for retailers already achieves much of the above.
– It also offers an SDK built on the posemesh standard: a flavor of DePin for AI perception.
For more color, see the full video below...
Want more XR insights and multimedia? ARtillery Intelligence offers an indexed and searchable library of XR intelligence known as ARtillery Pro. See more here.
– Think of this as a sort of mashup of computer vision, AI, Web3, and IoT.
– Expect to hear more about DePIN as spatial computing and AI continue to converge.
– Meanwhile, Auki's Convergent platform for retailers already achieves much of the above.
– It also offers an SDK built on the posemesh standard: a flavor of DePin for AI perception.
For more color, see the full video below...
Want more XR insights and multimedia? ARtillery Intelligence offers an indexed and searchable library of XR intelligence known as ARtillery Pro. See more here.
