We meet a wide range of needs from network design from the ground up to development support and consulting.
Sanei Hitec provides various algorithm design services tailored to customer needs, backed by insights gained from its unique artificial intelligence (AI) research. We can flexibly respond to customer requests, from full scratch design and development support to considerations for network lightweighting aimed at edge devices. ○ Full Scratch Design of Algorithms Designing the structure of neural networks and other algorithms from zero based on application needs. ○ Algorithm Refinement Enhancing the accuracy and lightweighting of algorithms based on the customer's initial ideas. ○ Collaborative Development Promoting the project together with the customer, from literature review to algorithm design and device implementation considerations. 【Main Business Achievements】 Development of segmentation algorithms, provision and support of learning environments (consulting) for major automotive-related companies, major electronics manufacturers, and image-related companies, etc. ○ Semantic Segmentation ○ Examples of other algorithms that can be designed Our company researches algorithms for image analysis and audio analysis, considering various applications and devices.
Inquire About This Product
basic information
○ Full Scratch Design of Algorithms Designing the structure of neural networks and other algorithms from scratch based on the intended use. ○ Algorithm Refinement Enhancing the accuracy and reducing the weight of algorithms based on the customer's initial ideas. ○ Collaborative Development Promoting the process together with the customer, from literature research to algorithm design and considerations for device implementation.
Applications/Examples of results
【Main Business Achievements】Development of segmentation algorithms, provision and support of learning environments (consulting) for major automotive companies, major electronics manufacturers, and image-related companies, etc. ○ Semantic Segmentation To enable the implementation of semantic segmentation algorithms on FPGA and MPU, we targeted PixelNet for lightweight optimization. - Reduction of input image size - Reduction of the number of input and output channels at each processing layer (e.g., input/output channels of Convolution layer: 512ch ⇒ 32ch) Through these methods, we achieved a 90% reduction in the number of parameters. Additionally, after lightweight optimization, we conducted retraining and evaluation, confirming that there was no significant difference in accuracy (mIoU) compared to before optimization. Furthermore, to implement on FPGA and MPU, we optimized the lightweight PixelNet for hardware. - Parameters converted to 16-bit fixed-point - Divided the processing of the Convolution layer in the later stage to occur immediately after Bilinear interpolation, etc. ○ Other examples of designable algorithms Our company researches algorithms for image analysis and speech analysis, considering various applications and devices.
catalog（1）
Download All Catalogs
Sanei Hitec Co., Ltd. Company Profile
Company information
Our company has built a track record as a professional group in LSI design, distinguishing ourselves as an independent design specialist firm by having an LSI design environment equivalent to that of our clients and being able to handle the entire process of LSI design, from circuit design to layout and verification, for both analog and digital. Over the past half-century since our founding, the environment surrounding the LSI industry has continued to change, with semiconductor needs constantly expanding across various fields such as home appliances, musical instruments, gaming consoles, smartphones, and automobiles. With the advent of the IoT era, we believe this change will accelerate even further. Therefore, we have established overseas subsidiaries to enhance our responsiveness with a view to global expansion. Furthermore, by maximizing the strengths of our professional group, where 90% of our employees are engineers, we are continuing our efforts toward the future, including a full-scale entry into artificial intelligence (AI) as a new pillar of our engineering business and promoting joint technology development with academia and government, aiming for further evolution as a company that can meet societal expectations and fulfill its responsibilities.
