Doc for version ddk-3.4.7.7

NPU Applications

Only New VIM4 supports NPU, you can check the version of your VIM4 here: VIM4 Versions

Linux

Only supports OpenCV4.

Get source code

You need to download the source code to your VIM4 board and compile on it.

Clone the NPU demo applications to somewhere, e.g. ~/workspace:

$ mkdir ~/workspace
$ cd ~/workspace
$ git clone https://github.com/khadas/vim4_npu_applications
$ cd vim4_npu_applications
$ ls
densenet_ctc  facenet  face_recognition  face_recognition_cap  retinaface  retinaface_cap  vgg16  yolov3  yolov3_cap  yolov7_tiny  yolov7_tiny_cap  yolov8n  yolov8n_cap

densenet_ctc, facenet, retinaface, yolov3, yolov7_tiny, yolov8n - Different demo for VIM4 NPU.

Please use convert tool version tag ddk-3.4.7.7 or higher.

Install dependences

$ sudo apt update
$ sudo apt install libopencv-dev python3-opencv cmake

Compile

Take yolov3 as an example, other demos are the same.

$ cd yolov3
$ mkdir build && cd build
$ cmake ..
$ make

Run

Yolov3

The demos of the Yolo series only support running under the desktop environment.

Before running, please compile the source code. The compilation method is same as Mobilenet V2.

Detect picture:

$ cd yolov3/build
$ ./yolov3 -p ../data/1080p.bmp  -m ../data/det_yolov3_int8.adla

Detection with camera:

$ cd yolov3_cap/build
$ ./yolov3_cap -m ../data/det_yolov3_int8.adla -d X -w 1920 -h 1080

x: the number for you camera device. such as /dev/video0, x is 0.

Table of Contents