Competition Rules Data and Evaluation Leaderboard FAQs
Using SOPHGO TPU Submission Requirements

Using SOPHGO TPU

1.  CV1800B/SG200x Series Development Board

 

The CV1800B/SG200x series development boards are open-source hardware designed based on the SOPHON CV1800B/SG200x processors, capable of running operating systems based on LINUX and RTOS. They provide professionals, industrial OEM enterprises, IoT enthusiasts, amateur hobbyists, DIYers, and creators with a reliable, low-cost, and highly functional platform. These chips have successfully been ported to more than ten operating systems, including OpenEuler, ArchLinux, Fedora, Debian, OpenWRT, FreeRTOS, AliOS, RT-Thread, and have adapted software stacks such as opencv-mobile, ncnn, SQLite, while also supporting over 50 types of sensors.

 

2. Radxa Fogwise 1684X Mini Microserver

The Radxa Fogwise 1684X Mini microserver is equipped with the smart computing processor BM1684X, boasting up to 32Tops INT8 peak computing power, 16GB LPDDR4X memory, and 64GB eMMC. It supports mainstream programming frameworks (including TensorFlow, Caffe, PyTorch, Paddle, ONNX, MXNet, Tengine, and DarkNet), has a complete and easy-to-use toolchain, and offers a low cost of algorithm migration. It supports edge deployment of super-large parameter models such as ChatGLM2-6B, AIGC, Llama2, SAM, Whisper, etc. Without the need for expensive hardware and complex configuration environments, it offers plug-and-play with just an Ethernet cable, significantly lowering the barrier to entry and ensuring data security and privacy. It can be widely used in education, cultural tourism, industry, parks, IoT devices, and more.

 

 

3. Shaolin Pi Development Board

 

The Shaolin Pi development board is a BM1684-based development platform with approximately 20 TOPS of computing power. It uses the BM1684 as the core device, being fully autonomous and controllable, offering powerful computing capabilities and multi-channel video encoding and decoding. It supports 3 mini-PCIe lanes and 4 USB ports, allowing for the expansion of various peripheral modules. It can be optimally configured according to scenario requirements, with the most reasonable cost, optimal energy consumption, and the best functional choices. The hardware ecosystem is rich, with a variety of connectable peripherals. It supports a rich software development ecosystem and mainstream deep learning frameworks. The "Shaolin Pi" core board can be expanded with screens, keyboards, mice, cameras, headphones, VR, and other devices. You can DIY a full-scenario edge computing workstation on the "Shaolin Pi" to practice various experiments. It can also be embedded into unmanned vehicles and drones to realize edge computing for mobile terminals.

 

 

Note: Each team can only choose one piece of hardware to participate in the competition.

 

  • Participation resource description

 

Due to the affordability of these development boards, they are only available for purchase independently:

  1. Purchase link for the Duo series development board: https://ewmmz.xetslk.com/s/7O7GA
  2. Purchase link for the LicheeRV Nano series development board: https://sipeed.com/licheerv-nano

 

  1. This hardware platform will offer 30 product testing opportunities. After registration is complete (no need to wait until the registration deadline), you can apply for an Airbox. The company will assess the situation and start sending out the boards in March. The company will organize engineers to provide training and answer questions for participants using the boards;
  2. Add the SOPHON Assistant on WeChat (see the end of the document), fill out the application form, and send the registration screenshot and contact information to the assistant;
  3. Teams applying for a development board need to pay a deposit of 400 RMB and ensure that the board is returned intact within 10 working days after the competition ends;
  4. If you receive a board but do not successfully participate in the competition, you must return the board as soon as possible and ensure it is undamaged, and the company will refund the deposit.

 

  1. After registration is complete (no need to wait until the registration deadline), you can apply for the Shaolin School development board. The company will assess the situation and start sending out the boards in March. The company will organize engineers to provide training and answer questions for participants using the boards;
  2. Contact the YingShi assistant (WeChat ID: STEAM_HZ), and send the registration screenshot and contact information to customer service. Customer service email: yingshi_mm@hw100k.com;
  3. Teams applying for a development board need to submit a deposit of 800 RMB and ensure that the board is returned intact within 10 working days after the competition ends;
  4. If you receive a board but do not successfully participate in the competition, you must return the board as soon as possible and ensure it is undamaged, and the company will refund the deposit.

Submission Requirements

  1. Design Report: Submit one electronic document (DOC/PDF) that includes analysis of the system's working principle and key technical principles (basic concepts, processing procedures, etc.), system architecture design (structural choices, module division, technical selection, interface description), detailed design and implementation (development board wiring diagram, software flowchart, key code analysis, etc.), and system testing and analysis (key technology execution time tests and analysis, whole machine execution time tests and analysis, system functions and other performance tests and analysis).
  2. Software Source Code: Submit a packaged file containing annotated source code, which includes both the unaccelerated source code and the source code that has been optimized for acceleration. The submitted code should be consistent with the detailed design described in the design report.
  3. Presentation PPT: Main work, innovation points, results, conclusions.
  4. Demonstration Video: Submit one demonstration video (MP4 format) with a duration of no more than 5 minutes; the video should mainly showcase the entire process of system implementation, the thought process behind acceleration optimization, and a comparison of effects before and after acceleration. The video explanation should be clear and complete, the demonstration should be smooth, and the data presentation should be convincing.
  5. Other aspects that are beneficial for showcasing the work: For example, photographs of the hardware.