Real-time maize seed sorting with deep learning and embedded systems
This solution utilizes embedded GPUs and FPGAs for efficient real-time maize seed sorting through deep learning. It offers a compact, cost-effective, and energy-efficient system capable of 360-degree imaging and pneumatic sorting, ideal for single-person operation.
Overview
This innovative maize seed sorting solution leverages the power of embedded systems and deep learning technologies to achieve real-time sorting efficiency. By integrating embedded GPUs and FPGAs, this system provides a compact, cost-effective, and energy-efficient alternative to traditional seed sorting methods. The device is designed to be portable, fitting in the trunk of a car, and can be operated by a single individual. It is particularly suitable for remote locations, making it a versatile tool for seed sorting applications.
Technical specifications
Core Features:
- Utilizes embedded GPUs and FPGAs for low-power, high-efficiency computing
- Employs CNN pruning and knowledge distillation for enhanced deep learning model performance
- Features a 360-degree imaging system using three RGB cameras
- Integrates a pneumatic ejector for effective sorting
- Designed with a vibrating bed, chute, and hopper for efficient seed feeding
- Capable of operation in remote locations due to its compact size and power efficiency
Technology readiness level
Currently at Technology Readiness Level 3, this solution has been validated in controlled environments and is poised for further testing and development. Collaboration with seed and computer vision specialists is underway to advance the system towards field deployment. Active partnerships are sought for hardware development, funding, and collaboration.
About South Dakota State University
South Dakota State University is a comprehensive public land‑grant research university serving a large statewide community. Industry engages through a university‑linked research and technology park that colocates tenant companies with faculty labs and pilot‑scale processing suites for de‑risking and demonstration. A statewide extension network and field sites provide real‑world environments for validation and market feedback, while shared‑use core facilities and tailored workforce pipelines connect talent and tools to corporate R&D. Research is supported by competitive federal funding from USDA and NSF, with additional awards from NIH and DOE, alongside state and industry partnerships. A dedicated technology transfer office manages IP and industry agreements, and incubator space at the research park supports startup formation and scale‑up.
Inquire about this partnering listing
Industry applications
delivered to your inbox.