💡 DSD_MNIST_Streamline 💡

Project Overview

The DSD_MNIST_Streamline project aims to design and implement a digital signal processing system utilizing the MNIST dataset. This project consists of 5 layers, with each layer structured with local_ctrl, PU, and temp_bram components.

Board : Zybo z7-20

Project Result (SDK)

Accuracy: 100%

Speed

Total time in seconds: 0.000075 + 0.000090 + 0.000090 + 0.000089 + 0.000091 + 0.000089 + 0.000089 + 0.000090 + 0.000089 + 0.000098 = 0.000890 seconds
Milliseconds (ms): 0.000890 seconds × 1000 = 0.890 ms
Nanoseconds (ns): 0.000890 seconds × 1,000,000,000 = 890,000 ns

CPU Speed = 890,000 ns

---

HW Accelerator Speed = 69.992 ns

"The FPGA hardware accelerator is approximately 12.72 times faster."

Project Structure

The project is composed of the following five layers:

1. Layer 1
2. Layer 2
3. Layer 3
4. Layer 4
5. Layer 5

Each layer contains the following components:

• local_ctrl: Local control module
• PU: Processing unit
• temp_bram: Temporary Block RAM storage

Component Descriptions by Layer

Layer 1

• local_ctrl: The local control module in the first layer receives input signals and forwards them to the processing unit.
• PU: The processing unit processes the input signals and stores the results in the temporary BRAM.
• temp_bram: Memory for temporarily storing processed data.

Layer 2 ~ Layer 5

Layers 2 to 5 have similar components to Layer 1, but each layer is optimized according to the characteristics of the data being processed.

Simulation Results

The simulation results for each layer are as follows:

• Layer 1 Simulation Result

• Layer 2 Simulation Result

• Layer 3 Simulation Result

• Layer 4 Simulation Result

• Layer 5 Simulation Result

  

Implementation Result

The implementation result is as follows:

• Implementation Result
  
  

Utilization

Timing

Power

Conclusion

The DSD_MNIST_Streamline project successfully implements an efficient digital signal processing system through its five-layer structure. The simulation and implementation results for each layer validate the system's performance. This project demonstrates high performance and efficiency in processing the MNIST dataset.

Contact

For additional inquiries, please contact us at:

• Email: opqrs0422@naver.com
• GitHub: GitHub Profile