Future Thinking on the use of Video in Retail - Panel Discussion

IP Video technology is playing an ever-increasing role in the safety of store colleagues, security and retail operations across the whole retail sector, stores, distribution and supply chain

Every year, our video working group gets to hear from the Founders and CEO's of leading edge video providers via a panel discussion, chaired by Professor Adrian Beck, with the questions this year to Pierre Racz (Genetec), Gerard Figols (i-PRO) and Martin Gren (Axis) prepared and shared in advance.

In this years call, the questions covered different aspects of video technology, its applications, and the challenges and opportunities it presents. Here are a few highlights of that discussion.

The Role of AI and Intelligent Automation

Pierre Racz, CEO of Genetec, shed light on his distinction between AI and Intelligent Automation. He emphasized that while true AI may still be a distant goal, Intelligent Automation, which combines human oversight with machine learning, can yield better outcomes in retail video technology. 

Expanding Video Use Cases

Gerard Figols, from i-PRO, highlighted how video technology extends beyond security to encompass various retail operations. This includes understanding customer traffic, optimising inventory, and personalized interactions. Cameras are no longer just cameras; they are IoT sensors that provide valuable insights to retailers.

Ethical Considerations of Facial Recognition

Facial recognition technology was a hot topic during the discussion. Panellists stressed the importance of ethical use cases, consent, and transparency and compliance. Pierre Racz cautioned against covert use and emphasised that satisfied customers can share their experiences, while dissatisfied customers have a far-reaching impact.

Future-Proofing Video Investments

To maximise flexibility and avoid vendor lock-in, the experts recommended adopting open architectures in such fast-moving space. Martin Gren highlighted the significance of backward and forward device compatibility when making infrastructure choices.

Video Analytics in Self-Service Retail

The discussion covered self-checkout and self-service trends and noted the increased use of video analytics to enhance efficiency while countering fraud. But again, the consensus was that although the AI tools are good, to get the best results, the systems need clear oversight from well-trained people.

Innovations on the Horizon

The experts agreed that the future of video analytics in retail will continue to evolve, focusing on factors like audio, image fidelity, storage efficiency, and analytic sophistication. They emphasised the need to integrate third-party analytics and sensors to gain deeper insights and maintain ethical practices.

In conclusion....

The insights shared during the session provided a valuable look ahead at the role of video technology in the retail sector and a chance for the retailers in our group to reflect on the implications and how this new information can help them shape a video strategy and investment that can be more future proof.

For a deeper understanding of the current and future landscape watch the full session or read a summary transcript below. .

Summary transcript

Please note this is not a verbatim transcript, rather an edited synopsis of many of the key points addressed. Please refer to the full video for verbatim quotes.

Colin Peacock: We have an excellent session planned for our video working group.

I should start with the usual reminders around antitrust laws – everyone please keep this top of mind in terms of what we should and shouldn’t say regarding competitive, proprietary information. So far we’ve successfully avoided jail for 25 years and I’d like to keep it that way. I have no plans to go to jail, how about you Adrian?

Adrian Beck: No.

Colin Peacock: For those less familiar, ECR is a global retail association that has been operating for over 25 years. We have a huge number of retailer members worldwide who participate across various activities and working groups. We also have excellent academic support, like Professor Adrian Beck from the University of Leicester.

Additionally, we receive research funding and grants from a few selected partners, including Genetec, and their leaders often attend these sessions to stay abreast of the latest problems facing retailers that can inspire their innovation pipeline.

Our group objectives focus on sharing video technology best practices, furthering video analytics, and exploring in-store applications beyond just security use cases. Today we’ll concentrate on the future. We have multiple retailers joining us and we’ll hear from several later. Adrian Beck will facilitate the discussion - we’ve provided him the questions to ask the panel via an earlier retail member brainstorm.

Adrian Beck: Thanks Colin. Hello everyone, great to see you all again – hard to believe it’s been a year already! As Colin mentioned, we recently met with regular participating retailers to understand their most pressing video-related questions when thinking ahead.

We have 12 questions here that we’ll work through as best we can over the next 55 minutes or so. Each speaker has seen these in advance, so with no further delay let’s dive right into the first question around AI...

Adrian Beck: Let's start with you Pierre. What does AI mean to your Genetec business and how is it changing your product and service approach?

Pierre Racz: OK, first of all the acronym AI stands for "Absolute Ignorance."

People often misconstrue crafty guessing for actual intelligence when the technology remains completely ignorant. That's not to say these tools are useless though.

In all seriousness, I was at a conference where Microsoft, Intel and Nvidia's research PhDs spoke. They mentioned we remain far from achieving the technological singularity or human-level AI.

What we actually have now are statistical inference engines.

These are great at surfacing past patterns for us, but cannot drive innovation or discover new things the way humans can.

In the context of retail crime, real intelligent criminals can also find ways to trick these defenses through adversarial inputs.

So in summary, true AI does not exist yet.

What does exist is Intelligent Automation - this means smart systems doing the heavy lifting but with humans still playing an essential oversight role in the loop. This produces the best outcomes in my opinion.

At Genetec, we leverage machine learning as an acceleration tool but ensure people stay involved so we don't propagate "Absolute Ignorance".

Adrian Beck: Can you provide some real examples of how this intelligent automation philosophy manifests in your solutions?

Pierre: Sure - take license plate recognition, which we have offered within Genetec since about 2002. We originally used an approach called support vector machines that required manually building feature extraction models for different plate fonts, jurisdictional contexts, etc - very labour intensive.

About 10 years ago we switched to deep neural networks so the system could automatically self-learn features instead. However, because it's still fundamentally ignorant despite some clever illusion of intelligence, we caught unusual cases like it using license plate bolt patterns rather than printed numbers to guess at states. No matter how advanced these tools get, we must keep human oversight as part of the quality control loop.

Adrian Beck: Gerard, what's your perspective on AI and its impacts for IPRO? Do you share some of Pierre's scepticism?

Gerard Figols: Yes the hype around AI is definitely immense...I agree it's not truly intelligent or independent as Pierre said. However, video devices are still getting smarter. But smarter doesn't mean they can function autonomously without human guidance.

As camera counts grow massively, operators face overload. These smarter systems can help them monitor more devices by providing more insightful data. We don't see cameras as just cameras anymore - they are IoT sensors.

Deep learning allows them to classify objects within scenes so users can find the most useful events faster. But at the end of the day there must be a human operator analysing scenarios and deciding appropriate responses using these AI-generated insights.

Fusing all the data from various sensors - like video cameras among other sources - remains imperative as well.

Adrian Beck: Martin from Axis - where do you net out regarding the role and evolution of AI in the video surveillance domain?

Martin Gren: We have incorporated AI and deep learning across our chipsets and entire camera line up for years now actually.

But we don't use it just to grab headlines through fancy demos - the priority is improving core functions like image processing.

We also believe strongly in a hybrid model where edge devices enhance lower-level capabilities but cloud platforms drive heavier analytical workloads leveraging metadata from the cameras.

AI definitely provides lasting advances but claims around human-like reasoning remain overstated from what we've seen behind the scenes. Thoughtfully constructive applications paired with human interpretation lead to the best outcomes in the real world rather than radical autonomy.

Adrian Beck: Let's continue walking through our industry questions. Let's move to our next question around the rising challenge of organized retail crime that so many retailers on our call have faced lately.

Pierre Racz: There was an interesting article about a company in the United States that got into legal trouble for crowdsourcing some video labelling, and one person was wrongly tagged as a human trafficker. That individual was stopped aggressively by police but eventually released without any charges after 40 minutes of very rough detention. They are now suing that video camera manufacturer. So it highlights the great responsibility involved when interpreting video data.

If you will label footage, be extremely accurate. Actual humans should be involved in analysis before unfounded assumptions spread downstream. You also need properly scoped usage policies governing who receives such sensitive information.

For example NATO (North Atlantic Treaty Organization) has varying levels of intelligence data access across member states based on agreements and laws. Just like restaurants safeguarding food quality through supplier standards, video data requires thoughtful handling as well.

Adrian Beck: Gerard from IPRO - what's your advice around seamlessly and securely sharing video data with law enforcement for criminal investigations?

Gerard Figols: Two crucial points jump out right away.

First is guaranteeing data integrity, so proper encryption to make certain the information remains intact in transit.

The second consideration is bandwidth constraints, because sharing all raw video feeds from many retailers could overwhelm police departments and limit their capability to react.

Instead of transferring entire video archives, it likely works better to share analysed incident metadata like alerts and insights so they know where to focus.

But adequate infrastructure would need to support strong security protections as per regulations like GDPR (General Data Protection Regulation).

Adrian Beck: Martin from Axis - what have you seen relative to external video data exchange needs?

Martin Gren: Honestly the most common method I still encounter is people filming screens with their mobile phones, which is obviously not ideal.

Proper encryption and auditing capabilities are absolute must-haves.

Regarding formats, we have perceived greater comfort progressing the industry norm towards H.264-encoded MP4 video clips as needed. But overall this isn't an area we actively facilitate as a manufacturer beyond guidance.

Adrian Beck: Let's combine questions 3 and 4 since they relate... Retailers are very interested in quantifying video's business value beyond a security foundation. So first, how might video use cases stretch deeper into retail operations? And second, which processes could be automated using video analytics? Pierre, over to you for initial thoughts.

Pierre Racz: It's debatable how innovative this is since some customers have already achieved major results in these areas before using techniques that could now spread more broadly.

For instance, leading retailers who have successfully reduced shrinkage well below industry averages routinely reposition mobile cameras over different store zones for marketing and operational analytics beyond just monitoring cash registers.

We're also exploring camera-equipped robots autonomously patrolling locations after hours to detect restocking needs or inventory anomalies. So video and analytics are already transforming activities in measurable ways when deliberately leveraged, though mainly by big chains presently. Making such capabilities more pervasively accessible can catalyse the next evolution.

Adrian Beck: What are you observing relative to expanding video use cases Gerard?

Gerard Figols: Various insights come into focus through creative camera placements and AI-based apps - understanding customer traffic flows and hot spots; queue management to boost service levels; optimizing inventory; personalized interactions with VIP shoppers through face recognition and more.

We have over 150 deep learning-capable cameras to enable this edge-based data extraction.

Adrian Beck: Martin, what non-security use cases are you encountering?

Martin Gren: Self-checkout stands out as a prominent arena actively using connected video to enhance loss prevention while increasing efficiency through less labour.

That said, we have reservations around techniques like blacklisting.

It's easy to overstep ethical boundaries, even inadvertently, in retail environments open to the general public.

Claims we see around assessing falls also seem unrealistic despite advanced analytics. 

Humans still prove better at contextual judgment for now.

Adrian Beck: Facial recognition persists as a lightning rod issue given the appetite retailers have for leveraging it to curb abuse towards staff, balanced by strong privacy concerns.

Pierre, you've been vocal about related risks in the past - where do you stand given the promotional capabilities but simultaneously scary connotations?

Pierre Racz: Companies should first be clear whether their intent is security or surveillance-driven advertising. Ethical use cases center on voluntarily enrolled participants who opted in fully informed of trade-offs. Any other covert identification breeds long-term distrust once uncovered.

Genetec previously had retail customers switch off facial recognition after overwhelmingly negative customer responses - it came across as creepy with minimal crime deterrence value.

The core lesson here, reinforced by Dave Carroll's viral United Breaks Guitars masterclass, is that satisfied customers may tell a few others about their experience whereas dissatisfied customers have reach to tell millions instantaneously.

Avoid anything that can be perceived as intrusive.

Adrian Beck: But doesn't that contradict with protecting staff from abusive repeat offenders for example? What's your take Gerard?

Gerard Figols: The technology capability definitely exists - it's more about how its applied.

As with Internet cookies, obtaining consumer consent remains imperative. We're seeing more airport security and check-in protocols offer facial recognition as an opt-in acceleration boon for travelers willing to enroll their biometrics.

When ethically deployed with proper data safeguards, it can enhance safety and operations. Geographies have unique regulations guiding appropriate usage as well that must stay centrally in focus.

Adrian Beck: Martin, thoughts from Axis?

Martin Gren: Some Middle Eastern regions mandate facial tracking in ways unacceptable elsewhere, like schools exploring it only to face legal injunctions.

Beyond confirming local laws, retailers must keenly evaluate risks to their brand reputation. No organisation wants to star in the next PR crisis video gone viral!

That said, for consenting users in restricted areas like warehouses, it provides value. Just exercise extreme diligence given societal sensitivities.

Pierre Racz: A European bank discouraged in-branch verbal abuse by informing volatile customers they remained visible through slightly blurred public view monitors.

Utter transparency combined with technology fusion for ethics can bolster environments substantially.

For example, Genetec solutions certified by the respected European Privacy Seal guarantee encrypted video auditing without explicit profiling.

Avoid anything covert yet pursue creative strategies promoting honourable behaviour.

Adrian Beck: Let's continue exploring other areas of interest that surfaced during our retailer round tables. The next logical discussion area based on retailer polling centres on future proofing costly video investments spanning years typically before refresh cycles...So what advice do you have for maximizing flexibility? Pierre, please kick us off.

Pierre Racz: It's "telling" how all of the technology partners on this call here embrace open architectures rather than vertical integration or custom stacks.

Retailers must avoid vendor lock-in scenarios yielding few options.

Open platforms demonstrate committed innovation for customers versus stagnant reliance on fragile monopolies.

The promise lies in continuously harnessing advances across technology shifts rather than just exploiting temporary competitive advantages.

Adrian Beck: How do you guide customers on strategic system design principles Gerard?

Gerard Figols: We wholly agree openness injects crucial upgrade flexibility paired with interoperability for component swapping.

On top of that consider full lifecycle total cost of ownership, not merely upfront capital assets.

One lens should be cyber risk - failing to implement adequate authentication and data encryption makes expenditures wasteful.

Regional security regulations have complex nuances as well - conformity helps future proof platforms.

Adrian Beck: Martin, how do you advocate retailers approach these sizable video infrastructure choices?

Martin Gren: We acknowledge technology continually progresses so client refresh actually drives our roadmap prioritization. Ensuring both forward and backward device compatibility minimizes disruption when modernising.

Beyond gear, systemic considerations like warranties, maintenance policies and sustainability standards matter for total cost.

Particularly retail chains face increasing pressure to demonstrate environmental awareness and social governance. We must exhibit those ourselves as suppliers too. 

But adjusting incrementally on integrated foundations enables scaling securely.

Adrian Beck: Self-checkout and self-service boom across categories, bringing both perks and perils like efficiency gains countered by fraud upticks. What's the video analytics outlook here from each of your vantage points?

Pierre Racz: Visible cameras historically tempered misbehaviour through passive peer pressure.

Anonymizing self checkouts now erode that effect.

While technology can help significantly, the right staffing models retain importance - rewarding personnel supporting customers across physical and digital commerce may provide one path.

Store managers would gain from analysing shrinkage metrics across various checkout hybrid strategies rather than just eliminating cashiers outright.

Gerard Figols: Video devices monitoring consumers certainly assist. But corralling machine sensor data like weight discrepancies provides powerful fusion.

Surfacing self imagery triggers subconscious social cues we leverage for gentle self-policing too.

Integrated analytics sift operational data flows to enable internal loss prevention teams appropriately.

No doubt self checkout usage will continue rising, so sensible safeguards and oversight stay vital.

Martin Gren: We have seen steep adoption for self checkout driven by tangible retailer savings from needing fewer front end staff. Yet effectively maintaining these at scale takes systems integration mastery - what works in small trials may degrade across hundreds of locations.

Tools like product recognition have reached new levels of precision to speed flows in assisted modes as well. Various data harnessing assistance through video and adjacent innovations can collectively improve manned and unmanned store experiences.

Adrian Beck: Our last question...what do each of you believe retail video analytics innovations on the horizon beyond the current state?

Pierre Racz: Top retailers approach this holistically, embedding technology within processes, infrastructure and talent strategies concurrently.

Perhaps checkout personnel could enjoy career growth graduating into specialised loss prevention roles versus feeling threatened. Humans and smart machines collaboratively outperform either in isolation. Consistently analysing the interplay across dimensions unlocks lasting progress.

Gerard Figols: Being open to third party analytics and sensors multiplies insights fed through shared data repositories.

Acting upon amplified intelligence distilled from fuse data requires culture and capability depth.

But meticulously aggregating observations from video, point of sale, inventory and more enables realization of bigger opportunities.

Martin Gren: Core fundamentals like audio, image fidelity, storage economy and analytic sophistication will all keep maturing thanks to ongoing advances.

As sentry perimeter notions dissolve under perceptive uniformity, hitherto unmonitored areas like parking and loading docks warrant attention too.

Customers expect retailers to exhibit digital sophistication with sustainable purpose.

We have shifted from merely selling standalone products to promoting ethical client outcomes via managed services spanning hardware through software. Priorities like transparency and accessibility go hand-in-hand with technology growth to serve and protect people both online and in physical spaces.

Colin Peacock: Adrian please provide any final comments before we conclude this thought provoking session.

Adrian Beck: I would like to thank our panel - Pierre, Gerard and Martin - for these open insightful dialogues exploring so many intriguing themes on the future of video in retail. Colin back to you for final remarks…

Colin Peacock: Thanks Adrian. And thanks indeed to all our speakers along with the audience for joining what has been a stimulating discussion as we wrap up this year and gear up for 2024. We have captured excellent notes from today's conversation to shape ongoing working group activities. Please reach out with any additional questions. Happy holidays everyone!