Examples

To automate the routine of staff walking the plant to read and record analogue meters by eye, reducing headcount needs and improving efficiency on the factory floor.

Without replacing any existing meters or equipment, an app simply pointed at a meter by an AI camera reads the value automatically — offered as an app on Actcast. Because the analysis completes on a device costing a few thousand yen, the cost stays manageable even across a large number of meters. Running the analysis on the edge also means factory footage never leaves the site.

To manage IoT devices on production lines securely and turn the data gathered on the floor into improvement activity — reducing operational burden and improving how work gets done.

When device management is split across departments and plants, security settings drift apart and software updates get missed. Managing general-purpose devices such as Raspberry Pi centrally on Actcast — with remote configuration changes and software updates — removes that fragmentation. Combined with CTC’s data-enablement service “D-Native,” the collected data is carried through to visualization and analysis.

To give signalling equipment — devices that announce a situation with light and sound — the ability to perceive the situation itself, so the equipment can decide when there is something worth announcing.

Detection results from the “ai cast” edge AI camera are wired directly into network-controlled signal towers. The first application handles detection of white canes and wheelchairs, announcing the arrival of someone who may need assistance immediately through light, sound and voice. Because the judgement completes on the edge, the notification happens locally without footage leaving the site.

ActcastAI

Gate One wants to visualize the advertising impact of its checkout-counter digital signage, “Famima TV,” installed at more than 10,000 stores nationwide, in order to properly assess advertising value.

We handle everything end to end — developing the AI technology, launching the edge AI camera system, and long-term ongoing operation across thousands of stores in all 47 prefectures — to perform the visual recognition analysis itself. To realize an AI camera system that must never stop and must never lose data, we build in stable operation from the OS level up, alongside remote operational monitoring, root-cause isolation when anomalies occur, and monitoring AI accuracy. We also protect privacy by extracting only viewer demographic attributes and viewing time on the edge, discarding the identifiable images on site rather than sending them to the cloud.

ActcastAI

Sogo & Seibu wanted visibility not only into customers who bought something, but into the number, demographics and movement of everyone who visited — so that floor initiatives and tenant strategy could be based on evidence.

We installed edge AI cameras in the ceilings of the sales floors to estimate visitor counts and attributes (gender, age group). In later phases we added ReID — vectorizing facial features to judge whether two sightings are the same person — so movement across floors could be followed. Because the AI analysis runs entirely on devices costing a few thousand yen, deploying many cameras across large floors does not inflate the cost. Footage is converted to metadata on the spot, and no personally identifiable information is sent to the cloud.

JR Hakata City runs OMO-format stores to widen the opportunities for a variety of businesses to open shop, and wanted to hand those tenants data they could use for marketing.

We installed edge AI cameras in the store to capture visitor counts and attributes (gender, age group), plus how long signage and displays were actually looked at. Because the AI analysis completes on a credit-card-sized, inexpensive device, the setup can be deployed quickly enough to suit a limited-run pop-up store.

To cut store operating costs and improve the shopping experience at the same time — removing overnight work such as swapping out promotional materials, and reducing waste.

An AI camera judges, on the spot, how long a customer lingers in front of a shelf and which shelf they reach toward. Identifying the position of the wrist turns “walked past the shelf” into “showed interest in that product,” and this is wired into the signage delivery system to switch what is displayed. The aim is to remove the job of replacing paper promotional materials altogether.

Car parks themselves remain largely undigitized, and parking data sits separately from the data of the retail facilities attached to them. The aim was to bring the two together into something usable for marketing.

Occupancy detection, licence plate recognition and vehicle orientation are handled alongside people counting and attribute analysis — all on the same edge AI camera. Being able to capture both vehicles and people on one platform is what makes joining parking and facility data possible. Running it on the low-cost “ai cast” device keeps the equipment cost viable even when deployed across many parking bays.

Staff training had become dependent on individual supervisors, with no objective view of how service was actually being delivered. Customer harassment at the counter was a chronic problem, but its causes could not be pinned down.

AI microphones at the service counter record conversations between customers and staff with speaker separation. Making the quality of service objectively understandable from audio data means the same data serves both training improvement and harassment countermeasures. The visible presence of a microphone also has a deterrent effect of its own. The edge AI automatically detects when speech starts and ends, so store staff need no training to use it — zero operational overhead.

Looking ahead to the arrival of the smart city, AISIN wanted to enter the AI camera field and bring a capable, extensible product to market early.

The Raspberry Pi general-purpose computer was paired with the “Hailo-8” AI accelerator chip, and Idein wrote the software that draws out its performance. Actcast then took on the remote management needed for large-scale operation, which meant AISIN did not have to build device control software from scratch. That division of labour is what made launch six months from the start of development possible.

Setting the timetable for the Toei Subway requires knowing accurately which station, which car, and how many people are boarding and alighting. But surveys were carried out by staff observing in person, which limited the places, times and cars that could be measured.

Edge AI cameras on the platform measure the number of passengers inside the car and the numbers boarding and alighting while it is stopped. Because footage is never sent to the cloud and only counts are extracted on the spot, the setup can stay in place in a public space. Site-specific conditions — such as UV-blocking window film affecting analysis accuracy — were established by installing and testing rather than assumed.

To make smart buildings widespread. The obstacle was that each piece of equipment required its own development and operation, leaving the burden of managing edge devices customer by customer and building by building.

The “DX-Core” building OS and Actcast are integrated platform to platform, shifting edge device management onto Actcast. That removes the need to manage large numbers of devices separately in every building, and lets the integration itself be assembled faster and in a better shape. Actcast already had a track record of remotely operating AI cameras dispersed across large spaces such as commercial facilities, and that operational capability carries straight over to buildings.

In a large commercial complex, there was no accurate, real-time picture of how visitors moved through the building. Capturing entries and exits per floor would make time-of-day congestion and imbalance between floors visible, and give initiatives something to be judged against.

Within a framework that controls and manages cameras across the whole area centrally, we handle the AI image analysis. Being able to add analysis while keeping the existing camera network in place is a precondition at this scale. What is captured is statistics on counts and movement — the setup does not record individuals.