Leela Platform combines visual sensors with AI algorithms to provide continuous time and motion studies of activities involved in manufacturing. Our manufacturing intelligence software measures cycle times and value- vs. non-value add work time, compares performance across shifts and locations, and helps optimize staff redeployment due to labor or supply shortages or changes in processes. Leela Platform can identify safety, PPE, and ergonomic hazards, as well as capture best practices for use in training. The software provides a visual ground truth that can be used to compare against the observations of sensor analytics and quality detection solutions. All this adds up to reduced waste and higher capacity.

The Leela Viewer dashboard provides web and mobile views of key metrics for each workstation and production line. Statistics such as cycle time and value vs. non-value add time are expressed in easy to comprehend graphics. You can drill down into video clips with full VCR control and view time-series charts that display the duration of specific tasks of each workstation, line, or shift.

The greatest ROI is typically achieved in manufacturing sites that are high-touch (lots of people) and/or high-mix (lots of production changes). Yet, we can deliver ROI in any industrial operation where human workers play a role, including logistics. Other applications include measuring human work to assist in ERP or HR studies or providing baseline data to inform a transition to additional automation. Leela Platform is especially useful during labor shortages, letting you identify bottlenecks and inefficiencies and increase productivity despite static staffing.

Our SaaS cloud platform is available on a per camera subscription basis, letting you affordably trial the software and start small before expanding. We can set up in a day or two with standard, Ethernet connected cameras, and train the AI in as little as a week, based on 3-10 smartphone recorded video clips per camera station. The set-up time depends on the complexity of the recorded tasks and number of cameras. Typically, you can start generating actionable analytics within one to three weeks.

Facial blurring is the default. Options include full body blurring and the implementation of blurring at the initial point of capture and processing, thereby making facial data unrecoverable. There is also an option to disable video storage and review, in which case you work solely with aggregate data. In the future, we plan to offer an option to convert video into animations, which will later form the basis for digital twin simulations.

Leela Platform has been certified for the SOC 2 Type II security framework to ensure the highest level of protection against cyber-attacks. The video is owned by the customers. As a default, video is archived on our cloud servers for only 3 months, aside from clips saved by the customer. The cut-off is determined by the customers. Per special request, we can set up edge servers to keep video storage entirely on site.

Typically, Leela Platform is used by industrial engineers, continuous improvement officers, and line supervisors. Yet, the software achieves its greatest value when the Leela Viewer dashboard is shared throughout the organization, enabling insights from the front office to the front line. The dashboard supplies a common, ground truth reference to enable cooperation and improvement throughout the organization.

If you have a very highly automated factory with a lot of sensor analytics and related software, you may not need Leela Platform. Yet, sensor analytics are limited to monitoring equipment and cannot effectively analyze the steps in between that are performed by human operators. Also, many factories use legacy equipment that cannot easily or effectively be retrofitted with sensors. A MES requires operator input to record key steps, a task that is often skipped and is prone to error. Leela Platform can augment visual product inspection by spotting macro-level issues that affect quality. Our software can capture the processes that are missed by all these solutions: production tasks involving humans. It can also track and measure the use of tools, equipment, robots, transport carts, parts, products, and more.

We excel at rapid AI training (and retraining) and in accurately capturing and teasing out trends from complex visual data involving humans. We require 100x less data and 10x less time than typical solutions. These advantages are enabled by our hybrid causal/neural network AI engine, which was developed at MIT AI Lab. Our hybrid AI platform adds commonsense reasoning and cause and effect analysis to traditional neural network pattern recognition. This enables the AI to learn on its own, quickly grasp new concepts, explain its reasoning, and more accurately translate unstructured visual data into actionable insights.

We do not do video surveillance, and we perform visual product inspection only on the macro level. We do not currently do serial number scanning and tracking of individual units, but our time-coded visual data can be integrated via APIs with a product tracking solution. We offer near real-time response: Depending on networking bandwidth and the complexity of the scene, our feedback time for alerts ranges from 5 to 15 minutes, which is usually enough to detect production lines problems before they turn costly. At extra cost, we can reduce feedback time to as little as 1 minute.
You can start with a single camera and expand incrementally. We can fuse data from multiple cameras on a single workstation and integrate and compare visual data from up to a 100 cameras across multiple sites.
Leela Platform is highly customizable thanks to an AI that can grasp concepts in context and explain its behavior. How much is done by our staff vs. yours is up to you. Our customers can specify new activities to track by highlighting the area of interest on the video and adding new labels. An API is available to integrate Leela Platform’s visual analytics with IIoT, MES, ERP, and other enterprise software. Yet, the software can also provide significant value on a stand-alone basis.