Kneron Model Zoo Inference Example - Detection(Yolov5)

PLUS Inference Example with NEF model trained from https://doc.kneron.com/docs/#model_training/object_detection_yolov5/


Check - tutorial ( location: /workspace/ai_training/detection/yolov5/tutorial/tutorial.ipynb )

in

to learn how to train a the example model, and check

to learn how to convert the trained model to NEF model

Note: We built and run the examples below under OS Windows 10 (19041.1052), Keil uVision5 (5.27.1.0), MSYS2 MinGW 64-bit (20210419), and cmake version 3.20.2.


How to:

step 1. read document at Getting_start and make sure PLUS is ready

step 2-1. run KL520 example:

    cd build/bin/
    ./kl520_kn-model-zoo_generic_inference_post_yolov5 

step 2-2. run KL720 example:

    cd build/bin/
    ./kl720_kn-model-zoo_generic_inference_post_yolov5 

Detail Infomation:

Platform backbone FPS(npu only) Input format Preprocess Normalize
KL520 YOLOv5s 4.91429 RGB565 auto (on device) Kneron mode
KL720 YOLOv5s 24.4114 RGB565 auto (on device) Kneron mode