Archive by year
DMP, going into full scale in professional services for safe driving support system development
Systemizing the necessary functions into “ZIA™ SAFE” for realizing the driver monitoring, Hiyari-Hatto (near-miss) detection,
and tailgating detection systems using drive recorders (dashboard cameras)
Digital Media Professionals Inc. (Headquarters: Nakano-ku, Tokyo, President & CEO Tatsuo Yamamoto, hereinafter referred to as DMP) has developed AI platform “ZIA™ SAFE” specialized for safe driving based on ZIA™ Classifier which has already been proven as AI image recognition software. DMP will accelerate the provision of professional services that help customers develop safe driving support systems using drive recorders.
In recent years, as tailgating and dangerous driving by the elderly have become social problems, there has been an increasing momentum of developing safe driving support systems such as driver monitoring, near-miss detection and tailgating detection, by using drive recorders in the automobile and insurance industries.
DMP helps customers establish advanced safe driving support systems using AI in a short period of time through professional services with the newly developed “ZIA™ SAFE”.
DMP also helps the establishment of services that meet new needs related to CASE*1 and MaaS*2 through “ZIA™ SAFE”, such as collection of various road area information, traffic conditions and real-time analysis, by using drive recorders.
“ZIA™ SAFE” has evolved from ZIA™ Classifier which has been steadily adopted for drive recorder automatic analysis, automatic / autonomous driving, and medical applications since its launch in 2016, in order to realize safe driving support systems by systemizing necessary functions like object recognition, scene understanding, privacy protection, etc. into a collection of independent modules (building blocks). DMP can respond flexibly and quickly to the specific requirements of various customers for safe driving support system development and service establishment by combining the proven “ZIA™ SAFE” modules.
“ZIA™ SAFE” can be operated with different processors made by such as DMP, Intel, NVIDIA, Qualcomm and their combinations. As part of its professional services, DMP not only performs image data annotations*3, training of learning data, and UI*4 development, but also helps development of AI hardware-level acceleration technology and customer-specific hardware design based on its GPU development technology, a unique strength that DMP has accumulated over the years. Further optimization of system performance and power can be attained by utilizing these hardware related services.
Professional service for safe driving support systems by ZIA™ SAFE
Examples of adoption by customers in the mobility field
- Advanced safe driving system: AI drive recorder (Sumitomo Mitsui Auto Service Co., Ltd., Denso Ten Limited, etc.)
- Providing dangerous driving behavior detection technology that detects sudden braking, looking aside, falling asleep, lack of inter-vehicle distance, etc. based on video from in-vehicle cameras, various sensors information, and map information
- Automatic / autonomous driving (automobile equipment manufacturers, etc.)
- Providing technology for recognizing objects such as traffic signs, lanes, and road signs by deep learning from images captured by dashboard cameras
- Providing technologies such as SLAM*5 for spatial location information and Path Planning for route determination
- Optimizing the above technology model and implementing on GPU and FPGA
- SoC internal design (AI accelerator / image input processing (ISP), etc.)
Professional service provision process
- Survey and planning
By introducing DMP’s mobility technology and interviewing customers’ purposes of using AI, DMP clarifies what customers want to achieve by using AI and proposes the best way to proceed.
- PoC*6 development
After clarifying the purposes of AI utilization, DMP collects actual data from customers, or collects and creates necessary data, and performs annotations, selection of the optimal combination of ZIA™ SAFE building blocks, and evaluation and tuning of learning and accuracy. DMP usually uses CPU, GPU, or FPGA as the hardware platform for PoC development.
- System development
If the AI utilization effects can be expected in the PoC development, DMP performs development to implement AI model in the actual target system. The AI model is optimized and implemented according to the customers’ requirement of performance and power and the hardware platform (CPU, GPU, FPGA, etc.) to be used. DMP’s strength is to be able to offer technology from reducing the weight of AI models to optimizing GPUs, DSPs, and AI accelerators used by hardware platforms, or logic designing for SoC / FPGA for hardware acceleration.
Since the accuracy of the AI model changes due to changes in the external environment (changes in usage conditions, etc.), DMP offers periodic relearning and optimization. In addition, DMP optimizes the AI model according to changes in the customers’ target platform.
DMP will continue to research and develop cutting-edge technologies related to AI / deep learning technology and contribute to the intelligence of customers’ systems.
- *1 CASE
- A coined word that combines the initials of Connected, Autonomous, Shared, and Electric, and represents the current trends in the automotive industry.
- *2 MasS
- Abbreviation for Mobility as a Service. Next-generation transportation that integrates autonomous driving, AI, open data, etc. and integrates sharing services with conventional transportation and transportation methods.
- *3 Annotation
- Tagging of all forms of data such as text, voice, and images so that machine learning algorithms can recognize patterns.
- *4 UI
- Abbreviation for User Interface. An interface for exchanging information between users and products.
- *5 SALM
- Abbreviation for Simultaneous Localization and Mapping. Perform self-location estimation and map creation simultaneously from information obtained from various sensors.
- *6 PoC
- Abbreviation for Proof of Concept. Verification and trial about feasibility before introducing a new concept, theory or principle in full scale.