5-Layer Q8.8 Fixed-Point Streaming CNN Accelerator in Pure VHDL RTL on Artix-7 FPGA (Nexys 4) with Raspberry Pi 5 8 MHz SPI Co-Design (100% Bit-Exact Parity, 15x Cycle Reduction)
-
Updated
Oct 12, 2026 - TeX
5-Layer Q8.8 Fixed-Point Streaming CNN Accelerator in Pure VHDL RTL on Artix-7 FPGA (Nexys 4) with Raspberry Pi 5 8 MHz SPI Co-Design (100% Bit-Exact Parity, 15x Cycle Reduction)
To associate your repository with the hardware-for-ai topic, visit your repo's landing page and select "manage topics."