Skip to content

Latest commit

 

History

History
179 lines (98 loc) · 6.51 KB

Readmeori.md

File metadata and controls

179 lines (98 loc) · 6.51 KB
Raw

MAX78000 YOLOv1 Demo

Overview

The YOLOv1 Demo software demonstrates identification of a number of persons from their facial images using MAX78000 EVKit.

YOLOv1 Demo Software

Building Firmware

Navigate directory where the YOLOv1 demo software is located and build the project:

$ cd Examples/MAX78000/CNN/yolov1_demo
$ make distclean
$ make

Loading Firmware Image to Target

Connect a USB cable to CN1 (USB/PWR) and turn ON the power switch (SW1). Note the COM port (COM_PORT) of this connection from your system configuration.

Connect the PICO debug adapter to JH5 SWD header.

Load firmware image using OpenOCD as described in the SDK documentation. Make sure to remove PICO adapter once the firmware is loaded.

Host Application

Prerequisites

  • Python 3.6.9 or higher (tested for 3.6.9, 3.7.7 and 3.8.1)
  • NumPy (>1.18)
  • Scipy (1.4)
  • PyQt5 (5.9.2)
  • OpenCv (>3.4)
  • PySerial (>3.4)
  • MatplotLib (>3.2)
  • PyTorch (>1.3.1)
  • TorchVision (>0.5.0)

If an ai8x-synthesis virtual environment does not already exist, follow the instructions in the main repository before continuing.

Add the additional packages using:

(ai8x-synthesis) $ pip3 install -r requirements-faceid.txt

Running the Host Application

Navigate to directory demo and run the run_demo.py script:

$ cd Examples/MAX78000/CNN/yolov1_demo/demo
$ python run_demo.py -c <COM_PORT>

<COM_PORT> is the Windows serial port identifier (e.g., COM1) or a Linux or macOS device (e.g., /dev/tty.usbserial-D308XSRX)

When the demo window is open, it is possible to load images from disk or capture images from the PC web cam. Currently, the app database includes images for five female and five male celebrities.