Yolov3-tiny_3l on raspberry pi

[kalooy-e]

Can i use yolov3-tiny_3l on a raspberry ?I want to grocery products with a total of 50 classes. What is the best cfg should I use?

[stephanecharette]

I have 1 grocery store customer that uses a NN I built with yolov3-tiny on a RPI Zero. But I only have 12 classes, not 50. (https://www.ccoderun.ca/programming/ml/store.html) I don't know how the .cfg file yolov3-tiny_3l is different from yolov3-tiny, I've never looked at that first one.

We so need a web page to describe all the different .cfg files, and what they attempt to do differently from each other. I have no idea what they are, and I cannot be the only one.

(Edit to clarify: I didn't build it on the RPI0. But the final product runs on a RPI0.)

[kalooy-e]

Thank you so much for the reply. It helps me a lot

[stephanecharette]

We so need a web page to describe all the different .cfg files, and what they attempt to do differently from each other.

Started to document all the .cfg files: https://www.ccoderun.ca/DarkMark/Configuration.html

[ghost]

@stephanecharette so when i want to inference using yolo v4 on my raspberry pi:

• what steps to build yolo v4
• how to model to inference, etc/..

[AlexeyAB]

• by using Darknet, set OPENMP=1 in the Makefile, run make, download https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v3_optimal/yolov4.weights then run https://github.com/AlexeyAB/darknet#how-to-use-on-the-command-line ./darknet detector demo cfg/coco.data cfg/yolov4.cfg yolov4.weights test.mp4

./darknet detector demo cfg/coco.data cfg/yolov4.cfg yolov4.weights -c 0

./darknet detector demo cfg/coco.data cfg/yolov4.cfg yolov4.weights rtsp://login:pass@192.168.0.228:554

• by using OpenCV-dnn, compile OpenCV and use: https://docs.opencv.org/master/da/d9d/tutorial_dnn_yolo.html

  • c++: https://github.com/opencv/opencv/blob/master/samples/dnn/object_detection.cpp
  • python: https://github.com/opencv/opencv/blob/master/samples/dnn/object_detection.py

• or you can buy such Raspberry Pi with Intel neurochip Myriad X, it will run 10x times faster: https://shop.luxonis.com/products/depthai-rpi-compute-module-edition Compile OpenCV with OpenVINO backend: https://github.com/opencv/opencv/wiki/Intel's-Deep-Learning-Inference-Engine-backend

---

Read: https://github.com/AlexeyAB/darknet#yolo-v3-in-other-frameworks

[ghost]

@AlexeyAB Thanks for your hard work, i think this repo has found it's 1 man army in you