NEC is developing AI technology to digitize work tasks without the need to train videos from multiple cameras covering wide area work sites in advance and use them.

AI Video & Visuals


Tokyo, August 27, 2025 – (JCN Newswire) – NEC Corporation (NEC; TSE:6701) has developed AI technology that allows workers to recognize and digitize tasks, without pre-training and using on-site video, without on-site video, with on-site video, without pre-training and using local video from a wide range of cameras, including warehouses, factories and construction sites.

This newly developed technology can be installed immediately in an industrial workplace, contributing to visualization of the worksite that was previously impossible, thereby increasing productivity, optimizing talent allocation, and streamlining workflows. NEC plans to commercialize the technology by fiscal year 2026.

Although labor shortages in distributed warehouses, factories, construction sites, and other industrial workplaces have been strengthened, many processes that rely on physical labor remain. This is why there is a growing demand for optimizing personnel and work processes by visualizing working conditions and making more effective use of the limited workforce. Although technology already exists to recognize work tasks from videos, preparation takes a lot of time and effort to recognize specific tasks in the workplace, including on-site video data collection and training AI models.

Furthermore, to digitize work tasks across wide area work sites using multiple cameras, workers must be identified across the video of all cameras, and the perceived results of work tasks must be integrated for each worker. However, conventional technology has made it difficult to accurately distinguish workers wearing the same uniform and continue to identify the same person with multiple cameras without errors.

The features of NEC's newly developed technology include:

1. You can quickly install video work tasks by simply entering text

Using the Vision Language Model (VLM), NEC has developed AI technology that allows it to recognize a variety of work tasks without the need for pre-training or using video data. Recognition requires text input describing individual work tasks such as “getting packages from shelf” to select tasks and “pressing cart to transport items” to transport cart.

In the past, recognition of work tasks required video data collection and annotation, and AI model training. Furthermore, identifying related objects that workers interact or operate has proven traditionally difficult, and recognizing tasks from videos on industrial sites where various objects are mixed together has been a challenge.

This newly developed technology first (a) utilizes VLM in advance to analyze and extract features from text input describing individual working tasks. When analyzing videos, this technology (b) identifies related objects that workers interact or manipulate using their own AI model to capture relationships between people and objects, and uses VLM to extract functionality from images containing the objects identified as workers. By comparing and matching the functions extracted in (a) and (b), working tasks can be recognized from text input alone.

2. Contribute to the optimization of wide area work sites through digitalization of worker movements and tasks

Multiple camera use contributes to optimizing on-site work by identifying workers who are moving across the site without relying on clothing or other visual characteristics, and by digitizing each worker's work tasks over a long period of time across the workplace.

This technique allows the same person in multiple cameras with high accuracy by estimating the location of individual workers in the digital twin space (3D coordinates) by referring to the position of workers displayed in each camera image (2D coordinates) and measuring the proximity of the digital twin position and motion pattern. Additionally, the camera parameters required to estimate the location of workers (i.e. camera position and orientation) are automatically estimated, minimizing the time and effort required for on-site installation.

*NEC is developing image recognition technology to digitize various work activities

https://www.nec.com/en/press/202211/global_20221128_01.html

About NEC Corporation

NEC Corporation has established itself as a leader in the integration of IT and network technology, promoting a brand statement that “aligns a bright world.” NEC allows businesses and communities to adapt to the rapid changes occurring in both society and markets to provide social values ​​of safety, security, equity and efficiency, promoting a more sustainable world with the fullest potential for everyone. For more information, please visit NEC at https://www.nec.com.

Source: NEC Corporation

Copyright 2025 JCN Newswire. Unauthorized reproduction is prohibited.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *