Skip to content

Latest commit

 

History

History
87 lines (65 loc) · 5.2 KB

README.md

File metadata and controls

87 lines (65 loc) · 5.2 KB
Raw

Rock 5 image

output image

A Rock 5 image with OpenCV, ncnn, TNN and NPU

License

Installation.

  • Get a 32 GB (minimal) SD card holding the image.
  • Download the RockPi5.img.xz image (2.8 GByte) from our Sync site.
  • Flash the image on the SD card with the Imager or balenaEtcher.
  • Insert the SD card in your Rock 5 and enjoy.
  • Username: rock
  • no password

Model performance benchmark(FPS)

All models, with C++ examples, can be found on the SD image.

demo model_name inputs_shape dtype Rock 5
yolov5 yolov5s_relu [1, 3, 640, 640] INT8 50.0
yolov5n [1, 3, 640, 640] INT8 58.8
yolov5s [1, 3, 640, 640] INT8 37.7
yolov5m [1, 3, 640, 640] INT8 16.2
yolov6 yolov6n [1, 3, 640, 640] INT8 63.0
yolov6s [1, 3, 640, 640] INT8 29.5
yolov6m [1, 3, 640, 640] INT8 15.4
yolov7 yolov7-tiny [1, 3, 640, 640] INT8 53.4
yolov7 [1, 3, 640, 640] INT8 9.4
yolov8 yolov8n [1, 3, 640, 640] INT8 53.1
yolov8s [1, 3, 640, 640] INT8 28.5
yolov8m [1, 3, 640, 640] INT8 12.1
yolox yolox_s [1, 3, 640, 640] INT8 30.0
yolox_m [1, 3, 640, 640] INT8 12.9
ppyoloe ppyoloe_s [1, 3, 640, 640] INT8 28.8
ppyoloe_m [1, 3, 640, 640] INT8 13.1
yolov5_seg yolov5n-seg [1, 3, 640, 640] INT8 9.4
yolov5s-seg [1, 3, 640, 640] INT8 7.8
yolov5m-seg [1, 3, 640, 640] INT8 6.1
yolov8_seg yolov8n-seg [1, 3, 640, 640] INT8 8.9
yolov8s-seg [1, 3, 640, 640] INT8 7.3
yolov8m-seg [1, 3, 640, 640] INT8 4.5
ppseg ppseg_lite_1024x512 [1, 3, 512, 512] INT8 27.5
RetinaFace RetinaFace_mobile320 [1, 3, 320, 320] INT8 243.6
RetinaFace_resnet50_320 [1, 3, 320, 320] INT8 43.4
PPOCR-Det ppocrv4_det [1, 3, 480, 480] INT8 31.5
PPOCR-Rec ppocrv4_rec [1, 3, 48, 320] FP16 35.7
  • Due to the pixel-wise filling and drawing, segmentation models are relatively slow

Tips.

  • If you need extra space delete the opencv and the opencv_contrib folder from the SD card. They are no longer needed since all libraries are stored in the /usr/ directory.
  • Use a tool like GParted sudo apt-get install gparted to expand the image to larger SD cards. We recommend a minimum of 64 GB. Deep learning requires a lot of space.
  • An example of YoloV5 running on the NPU (25 FPS) is included.

Pre-installed frameworks.

Thanks.

A more than special thanks to Stuart Naylor, who, ever so kindly, provided us the Rock Pi 5 for free.

output image

paypal