Panasonic will launch face recognition server software with deep learning technology

The face recognition server software WV-ASF950 and the face registration extension kit WV-ASFE951W will be available overseas in July 2018, and will be available in Japan in August.

Panasonic announced that it will release face recognition server software using deep learning technology in overseas in July 2018. The software will be available in Japan in August 2018.

Panasonic will launch face recognition server software with deep learning technology

This high-precision face recognition software uses the world's highest face recognition performance (*1) core engine to recognize faces that are difficult to recognize with traditional techniques, including face deflection to the left or right up to 45 degrees, up or towards The next 30 degrees and some of the faces that are covered by sunglasses.

In addition, the new software uses Intelligent Auto (iA) mode (*2), which automatically adjusts the camera settings to capture the best image for face recognition. If you use this software with Panasonic i-PRO EXTREME series network The camera is used in combination with the “Best Lens License Key” bundled with the software in the webcam to send only “Best Shots” to the server for face recognition. Panasonic core devices and faces The combination of recognition software maximizes the performance of the software core engine and achieves high-precision recognition. The company plans to add a new feature at the end of this year to identify faces that are partially obscured by masks, which is difficult to solve with traditional face recognition systems. The problem.

Moreover, when the software is used in combination with an iA-equipped camera, image analysis can be performed on the camera instead of on the server, so only the best image is sent to the server. This reduces server and network load and reduces overall system cost. Taking the example of connecting 10 or more webcams in the system, the system cost can be reduced by about 40-50% compared to conventional systems that do not use the "best lens" function.

Panasonic will continue to improve security-related products and provide a variety of solutions to meet the increasingly diverse and evolving customer needs, such as surveillance systems integrated management face recognition solutions. By providing these solutions, the company is committed to being a “full integrator” that contributes to the customer's frontline operations.

Panasonic Deep Learning Face Recognition Software has the following features:

1. High precision: - evaluated by NIST (IJB-A Face Challenge) as the world's highest precision face recognition engine
- iA function and "best lens" image function for optimal face recognition engine performance and high recognition accuracy

2. System cost reduction: only send "best lens" image, reduce server load and network load

3. System scalability:

- Register up to 30,000 faces (optional)

- Integrated management of i-PRO monitoring system: comprehensive management of facial recognition system and monitoring system using the same GUI

Note:

*1 In April 2017, the comparative test of the National Institute of Standards and Technology (NIST), one of the world's most authoritative organizations (IJB-A Face Verification Challenge Performance Report / IJB-A Face Recognition Challenge Performance Report) The product exhibits the highest level of face recognition performance.

*2 A feature that uses Panasonic's "LUMIX" consumer digital camera technology to capture the best images for face recognition. It automatically detects moving objects, motion speed, face and video in videos that are often difficult to see due to subject motion and backlighting. Light intensity (whether day or night), real-time optimization of settings

[Development background]

With the rapid spread of the Internet of Things and AI, the realization of various information interconnections to make human life more convenient has emerged in various industrial sectors. To this end, a new technology called "deep learning" has been applied. Panasonic has adopted this deep learning technology in its face recognition products for the security industry. This enables the successful development and commercialization of face recognition technology and successfully overcomes the difficulties of traditional technologies, such as recognizing face tilt, face changes due to age changes or partially obscured by sunglasses. Panasonic will use this product for personal identification solutions for video security systems in various environments, such as public facilities and entrance management monitoring.

[Features]

1. High precision:

Top face recognition engine evaluated by NIST (IJB-A face challenge)

- The deep learning technology used by the new software was developed by Panasonic and the National University of Singapore, which improves face recognition performance by up to 500% (*3) compared to traditional systems.

*3 Even if the face recognition error rate of the IJB-A face image database is set to 0.01, the bad recognition rate is as low as 20%.

- A unique algorithm that combines the machine learning method, deep learning, and the similarity calculation method for suppressing the error rate, which can be recognized in the case where the traditional face recognition technology is difficult to recognize, for example, tilting on the face (face to the left) Or deflect 45 degrees to the right, 30 degrees up or down), sunglasses or a mask (*4) partially obscures the face and age changes cause facial changes.

*4 Plans to be launched before the end of 2018

iA features and "best lens" images for optimal face recognition engine performance and high recognition accuracy

- The iA function allows the camera to automatically detect scenes and automatically optimize settings accordingly to improve the detectability of video images. The camera automatically detects moving objects, motion speed, face and light intensity (day/night/headlights) in a video that is often difficult to see due to subject motion and backlighting, and optimizes settings in real time to capture better quality video.

- The "Best Shot License Key" bundled with the software can be installed in the Panasonic i-PRO EXTREME series webcam (sold separately) to automatically automate multiple face images captured when someone walks past the camera Select an image that is suitable for face recognition and send only those selected images to the server. It is therefore possible to send high quality images suitable for face recognition and reduce server load.

2. System cost reduction:

Send only "best shot" images, reducing server load and network load

- In the conventional face recognition system, all captured images are sent to the server, after which the server performs face detection and face recognition, and the server collectively carries the data processing load. Since sending all images requires a large amount of bandwidth and saving these images requires high-capacity hard disk space, the system tends to be large.

- Panasonic system integrates this product with the iA-capable camera, which can be used for face detection by the camera and only sends the "best lens" image containing the face to the server, which will significantly reduce the network load, so no large bandwidth is required . Face recognition using the "best shot" image on the server reduces server load and hard drive capacity requirements and allows up to 20*5 webcams to be connected to a single server.

- Unlike traditional systems, this Panasonic system no longer needs to send data-intensive high-resolution images to the server. Provides high-precision face recognition while significantly reducing network traffic, transmission costs, and network construction costs.

The cost of using a Panasonic system that connects 10 or more webcams can be reduced by about 40-50% compared to conventional systems that do not use the "best lens" feature.

*5 The number of cameras that can be connected depends on the flow of people passing through the camera and the retention period of the face image.

3. System scalability:

Register up to 30,000 faces (sold separately)
Up to 10,000 faces can be registered in a simple batch using the standard face recognition server software (WV-ASF950). The optional Face Registration Extension Kit (WV-ASFE951W) can register up to 30,000 faces in large installations.

i-PRO monitoring system integrated management

By using the WV-ASM300 or WV-ASE231W client software to comprehensively manage the Panasonic i-PRO monitoring system, users can perform face detection, face recognition and recorded video tracking in the same GUI. Therefore, the special operations and management required by the conventional face recognition system are eliminated, and centralized operations are realized.

specification

-Face recognition server software

Product number

WV-ASF950

WV-ASF950W

Overview

Identify the number of faces: Up to 10,000 faces can be registered. With the Face Registration Extension Kit, the number of faces can be increased to 30,000.

Number of cameras: up to four cameras can be connected

With the existing camera extension kits WV-ASFE901 (x 1) and WV-ASFE904 (x 4), the number of cameras that can be connected can be expanded up to 20.

-Face registration extension kit

Product number

WV-ASFE951W

Overview

The number of standard face recognitions can be expanded from 10,000 to 20,000 (one license) or 30,000 (two licenses).

Engine Overhaul

Engine Overhaul,Black Spacer Bush,Ferrous Metal Bushing,Ferrous Metal Spacer

Jinan Guohua Green Power Equipment Co.,Ltd. , https://www.guohuagenerator.com