/PRZWT/BEIJING, April 09, 2024 - At a time when AI technology is disrupting all walks of life, ChatGPT has phenomenally promoted the application scenarios of generative AI in various industries.
On April 2, the 2024 Deschmann Innovation Technology Pre-Communication Conference was held in Beijing. Sang Shengwei, director of Deschmann Technology R & D Center, revealed two breakthrough technologies in the industry - GPTfinger and Longting Motor.
Among them, the exclusive GPTfinger of Deschmann smart lock is the first time in the industry to apply GPT technology to the fingerprint recognition function of smart locks. GPTfinger solves the pain point of insensitive fingerprint recognition of current smart locks for people with poor fingerprint quality, making it easy to use for more than 80% of people with poor fingerprints.
GPTfinger: Fingerprint Recognition Technology for Deschmann Generative Pretrained Model
As a leader in the smart lock industry, Deschmann Smart Lock pioneered the industry-exclusive GPT finger, breaking through technical limitations and applying GPT-4 technology to fingerprint recognition to create an affordable GPT for fingerprint locks.
It is understood that GPTfinger has four major technological breakthroughs.
First of all, GPT-4 model is huge, space and computing power resource requirements are high. In this regard, Deschmann uses model distillation to compress the space and improve the computing power: keep the fingerprint recognition related 1.20 million parameters, reduce the model space from TB level to MB level, and keep the fingerprint recognition 1 million-level graph to graph model training; increase 10 times the main frequency and 10 times the computing power, and enable the hardware-accelerated MPU to achieve resource expansion.
Secondly, Deschmann's industry pioneered the "spiral circle field technology", imitating the radar wave search spiral method, forming a spiral sequence of fingerprint features in any circular area, using GPT technology to learn, so as to build an accurate fingerprint feature context model, transforming fingerprint features into relational contexts for generative training and learning.
Thirdly, Deschmann uses GPT's mask occlusion generative training, which can randomly "occlude" part of the fingerprint features through the algorithm for training, so that Deschmann GPTfinger can accurately match even in the case of missing part of the fingerprint features, cope with fingerprint wear and tear, incomplete situation, without fear of missing part of the features.
Finally, in order to cope with the fingerprint changes caused by seasonal sensitivity, the Deschmann industry pioneered the use of Deep Convolutional Neural Network (CNN) big data homologous to GPT-4 to learn the evolution law of fingerprint features during seasonal changes, allowing Deschmann GPTfinger to predict the change of normal fingerprints before and after becoming peeling fingerprints, and achieve reliable matching between seasonal fingerprint changes and normal fingerprints.
In summary, Deschmann GPTfinger is not afraid of missing fingerprint features, not afraid of seasonal changes in fingerprints, not afraid of individual differences, and more secure. Not only improves the recognition speed and accuracy of smart locks, but also solves the problem that a small number of people with poor fingerprint quality are not sensitive to fingerprint recognition using smart locks, achieving a qualitative leap in user experience.
After the national lock product quality inspection center (Zhejiang) led the crowd test, the national notary agency notarized the whole process, Deschmann GPTfinger, people with poor fingerprints, more than 80% are easy to use.
The domestic smart lock demand structure is improving, and the technology iteration is leading the international market
In recent years, with the popularity of concepts such as smart security and smart home, the scale of smart door locks in China has continued to grow. It is understood that in 2023, daily consumption in the per capita consumption expenditure of residents will decline, but household consumption will show an improvement trend. Among them, the demand structure of the smart lock industry is improving, and the sales share of fully automatic smart locks has increased significantly. More and more consumers are willing to choose smarter and more convenient smart lock products.
Compared with developed countries such as Japan, South Korea, Europe and the United States, China's smart lock market penetration rate still has a large room for improvement, and the market scale is still expected to continue to grow; at the same time, China's smart lock technology level has led the international peers for one to two generations, and the domestic smart lock industry continues to emerge. Amazing black technology, such as biometric technology that tends to be multi-modal integration, and sentinel cat eye system that shoots 24 hours a day.
Deschmann released two major industry breakthrough technologies this time, successfully applying GPT-4 technology to the fingerprint recognition function of smart locks, and breaking through the limits of domestic motor technology, which will undoubtedly lead the domestic smart lock technology into a new era.
Exploring "Unseen Technology", Deschmann's highly anticipated launch
Deschmann's exploration in the field of smart lock technology has never stopped. As an industry leader, Deschmann always adheres to independent innovation and continues to lead the industry to pioneer in "unseen technology". As of now, Deschmann has 200 + patents and software copyrights, including fingerprint head, main board, management software, algorithms, molds, lock body, and design are all independent intellectual property rights; with 2 million + own/existing production capacity and supply chain, professional smart lock manufacturing management team and engineer team, and quality controllability guarantee system for large-scale production. Deschmann's emphasis and investment in iterative upgrades of core technologies enable it to continuously launch industry-leading results and meet the growing needs of users.
It is reported that on April 11, Deschmann will hold the 2024 global new product launch conference with the theme "Explore Unseen". In addition to GPTfinger and high-torque Longting motors, Deschmann will also release other leading technologies and release new blockbuster products equipped with breakthrough technologies, which will undoubtedly refresh the industry's awareness again.
