Product Info
I replaced a different company's GPU with SiMa.ai! [Overview]
SiMa.ai (hereinafter referred to as SiMa) is a device with strengths in edge AI, rivaling the performance of Alpha Inc., a company that provides GPUs (graphics processing units). There are also voices expressing interest in actually using SiMa's MLSoC (Machine Learning System-on-Chip). In this article, we will introduce how to port a model that was running on Alpha Inc.'s GPU to SiMa and actually run it.
Overall Table of Contents
The transplant procedure is as follows:
Step 1: Overview
- Overview
Step-2: Preparation
Step 3: Environment setup
Step 4: Implementation
Goal of this chapter "Overview"
- Understand the ease of development and porting using SiMa
- Understand the preparations for evaluation
- Understand the actual work procedure
- Identify points where you may have trouble
- Find tools that can further reduce labor hours
Image after transplantation is complete
Alpha company video
SiMa Video
This time, we ported a system that uses YoloV7 to track people's movements and perform heat map analysis of their stay status. Both systems use development kits provided by each manufacturer to analyze existing videos. After the analysis, people are detected in green frames, and their movement and stay status are displayed on a heat map that changes from blue to red.
*There may be some differences in smoothness depending on the display application.
The benefits of transplantation
- "Low power consumption"
- "Fanless"
- "High-performance processing per watt"
*For details, please refer to the Preparation section, Environment Setup sections 1-3, and Implementation section.
Thoughts on transplantation
The actual replacement work was carried out by our partner company, MonoStruct.
We interviewed MonoStruct about their thoughts on the future of SiMa through the porting process.
Q: Please tell us your impressions when you received the service!
I regularly work on developing cameras that utilize AI, and I felt that SiMa would be extremely effective for this purpose as well.
Taking advantage of its low power consumption, it seems like it could be used in devices that are operated in very restrictive environments, which is appealing to me.
Palette is provided free of charge and is an easy-to-use development environment for developers, which makes it attractive in that respect as well.
Q: How long did the actual schedule take?
Overall Schedule
Manual Check
50%
environment construction
20%
Application Creation
30%
The replacement was completed within three weeks.
Comment from Shinko Shoji
So it's possible to implant the device in just three weeks.
Q: What was the most difficult part?
Checking the manual, which accounts for a large proportion of the work, was very difficult.
Because SiMa is an American company, the documentation is mainly in English, and half of the work is spent on having to check multiple pages of the manual to understand the details of a function.
I think it's not that difficult as long as you can cover everything in the manual!
Comment from Shinko Shoji
If the manual is easy to read, it means that even more labor can be reduced.
thank you very much!
Towards further labor reduction
The interviews revealed that reading the manual was an issue.
Possible reasons for this include the fact that it is in English and the display style is unfamiliar to many people.
OpenAI's customizable "GPTs" can solve this problem. GPTs allows users to customize ChatGPT for specific purposes and generate their own AI tools. Please take advantage of this to further reduce development time!
summary
Through this chapter, you can see that it is possible to successfully port Alpha's GPU to SiMa. In the development of SiMa, it is possible to reuse the trained models used to operate Alpha's products. In addition, SiMa can achieve features not found in Alpha's GPU, such as "low power consumption," "fanless," and "high performance processing per 1W."
After this, we will introduce the porting procedure in detail, dividing it into preparation, environment setup parts 1-3, and implementation.
Company Profile
- Company Name
- SiMa Technologies, Inc.
- Established
- 2018
- location
- America (Silicon Valley)
- business
- Providing SoC and SDK specialized for image processing edge AI
Inquiries regarding this product
Telephone enquiries
Solution Technology Department 3
03-6361-8095Reception hours: Weekdays from 9:00am to 5:00pm
Contact by email
Inquiry Form
We deliver email newsletters that provide useful information such as the latest news and industry trends. Subscription is free and easy, so please sign up!
If you have any questions or comments about Shinko Shoji, or would like to discuss our products or services, please feel free to contact us using the inquiry form below.
Comment from Shinko Shoji
If a development environment you're using for the first time is difficult to use, it can cause problems in your development!
The fact that such a development environment is provided free of charge is certainly one of its attractions!